[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"similar-BeehiveInnovations--pal-mcp-server":3,"tool-BeehiveInnovations--pal-mcp-server":61},[4,18,26,36,44,53],{"id":5,"name":6,"github_repo":7,"description_zh":8,"stars":9,"difficulty_score":10,"last_commit_at":11,"category_tags":12,"status":17},4358,"openclaw","openclaw\u002Fopenclaw","OpenClaw 是一款专为个人打造的本地化 AI 助手，旨在让你在自己的设备上拥有完全可控的智能伙伴。它打破了传统 AI 助手局限于特定网页或应用的束缚，能够直接接入你日常使用的各类通讯渠道，包括微信、WhatsApp、Telegram、Discord、iMessage 等数十种平台。无论你在哪个聊天软件中发送消息，OpenClaw 都能即时响应，甚至支持在 macOS、iOS 和 Android 设备上进行语音交互，并提供实时的画布渲染功能供你操控。\n\n这款工具主要解决了用户对数据隐私、响应速度以及“始终在线”体验的需求。通过将 AI 部署在本地，用户无需依赖云端服务即可享受快速、私密的智能辅助，真正实现了“你的数据，你做主”。其独特的技术亮点在于强大的网关架构，将控制平面与核心助手分离，确保跨平台通信的流畅性与扩展性。\n\nOpenClaw 非常适合希望构建个性化工作流的技术爱好者、开发者，以及注重隐私保护且不愿被单一生态绑定的普通用户。只要具备基础的终端操作能力（支持 macOS、Linux 及 Windows WSL2），即可通过简单的命令行引导完成部署。如果你渴望拥有一个懂你",349277,3,"2026-04-06T06:32:30",[13,14,15,16],"Agent","开发框架","图像","数据工具","ready",{"id":19,"name":20,"github_repo":21,"description_zh":22,"stars":23,"difficulty_score":10,"last_commit_at":24,"category_tags":25,"status":17},3808,"stable-diffusion-webui","AUTOMATIC1111\u002Fstable-diffusion-webui","stable-diffusion-webui 是一个基于 Gradio 构建的网页版操作界面，旨在让用户能够轻松地在本地运行和使用强大的 Stable Diffusion 图像生成模型。它解决了原始模型依赖命令行、操作门槛高且功能分散的痛点，将复杂的 AI 绘图流程整合进一个直观易用的图形化平台。\n\n无论是希望快速上手的普通创作者、需要精细控制画面细节的设计师，还是想要深入探索模型潜力的开发者与研究人员，都能从中获益。其核心亮点在于极高的功能丰富度：不仅支持文生图、图生图、局部重绘（Inpainting）和外绘（Outpainting）等基础模式，还独创了注意力机制调整、提示词矩阵、负向提示词以及“高清修复”等高级功能。此外，它内置了 GFPGAN 和 CodeFormer 等人脸修复工具，支持多种神经网络放大算法，并允许用户通过插件系统无限扩展能力。即使是显存有限的设备，stable-diffusion-webui 也提供了相应的优化选项，让高质量的 AI 艺术创作变得触手可及。",162132,"2026-04-05T11:01:52",[14,15,13],{"id":27,"name":28,"github_repo":29,"description_zh":30,"stars":31,"difficulty_score":32,"last_commit_at":33,"category_tags":34,"status":17},1381,"everything-claude-code","affaan-m\u002Feverything-claude-code","everything-claude-code 是一套专为 AI 编程助手（如 Claude Code、Codex、Cursor 等）打造的高性能优化系统。它不仅仅是一组配置文件，而是一个经过长期实战打磨的完整框架，旨在解决 AI 代理在实际开发中面临的效率低下、记忆丢失、安全隐患及缺乏持续学习能力等核心痛点。\n\n通过引入技能模块化、直觉增强、记忆持久化机制以及内置的安全扫描功能，everything-claude-code 能显著提升 AI 在复杂任务中的表现，帮助开发者构建更稳定、更智能的生产级 AI 代理。其独特的“研究优先”开发理念和针对 Token 消耗的优化策略，使得模型响应更快、成本更低，同时有效防御潜在的攻击向量。\n\n这套工具特别适合软件开发者、AI 研究人员以及希望深度定制 AI 工作流的技术团队使用。无论您是在构建大型代码库，还是需要 AI 协助进行安全审计与自动化测试，everything-claude-code 都能提供强大的底层支持。作为一个曾荣获 Anthropic 黑客大奖的开源项目，它融合了多语言支持与丰富的实战钩子（hooks），让 AI 真正成长为懂上",151918,2,"2026-04-12T11:33:05",[14,13,35],"语言模型",{"id":37,"name":38,"github_repo":39,"description_zh":40,"stars":41,"difficulty_score":32,"last_commit_at":42,"category_tags":43,"status":17},2271,"ComfyUI","Comfy-Org\u002FComfyUI","ComfyUI 是一款功能强大且高度模块化的视觉 AI 引擎，专为设计和执行复杂的 Stable Diffusion 图像生成流程而打造。它摒弃了传统的代码编写模式，采用直观的节点式流程图界面，让用户通过连接不同的功能模块即可构建个性化的生成管线。\n\n这一设计巧妙解决了高级 AI 绘图工作流配置复杂、灵活性不足的痛点。用户无需具备编程背景，也能自由组合模型、调整参数并实时预览效果，轻松实现从基础文生图到多步骤高清修复等各类复杂任务。ComfyUI 拥有极佳的兼容性，不仅支持 Windows、macOS 和 Linux 全平台，还广泛适配 NVIDIA、AMD、Intel 及苹果 Silicon 等多种硬件架构，并率先支持 SDXL、Flux、SD3 等前沿模型。\n\n无论是希望深入探索算法潜力的研究人员和开发者，还是追求极致创作自由度的设计师与资深 AI 绘画爱好者，ComfyUI 都能提供强大的支持。其独特的模块化架构允许社区不断扩展新功能，使其成为当前最灵活、生态最丰富的开源扩散模型工具之一，帮助用户将创意高效转化为现实。",108322,"2026-04-10T11:39:34",[14,15,13],{"id":45,"name":46,"github_repo":47,"description_zh":48,"stars":49,"difficulty_score":32,"last_commit_at":50,"category_tags":51,"status":17},6121,"gemini-cli","google-gemini\u002Fgemini-cli","gemini-cli 是一款由谷歌推出的开源 AI 命令行工具，它将强大的 Gemini 大模型能力直接集成到用户的终端环境中。对于习惯在命令行工作的开发者而言，它提供了一条从输入提示词到获取模型响应的最短路径，无需切换窗口即可享受智能辅助。\n\n这款工具主要解决了开发过程中频繁上下文切换的痛点，让用户能在熟悉的终端界面内直接完成代码理解、生成、调试以及自动化运维任务。无论是查询大型代码库、根据草图生成应用，还是执行复杂的 Git 操作，gemini-cli 都能通过自然语言指令高效处理。\n\n它特别适合广大软件工程师、DevOps 人员及技术研究人员使用。其核心亮点包括支持高达 100 万 token 的超长上下文窗口，具备出色的逻辑推理能力；内置 Google 搜索、文件操作及 Shell 命令执行等实用工具；更独特的是，它支持 MCP（模型上下文协议），允许用户灵活扩展自定义集成，连接如图像生成等外部能力。此外，个人谷歌账号即可享受免费的额度支持，且项目基于 Apache 2.0 协议完全开源，是提升终端工作效率的理想助手。",100752,"2026-04-10T01:20:03",[52,13,15,14],"插件",{"id":54,"name":55,"github_repo":56,"description_zh":57,"stars":58,"difficulty_score":32,"last_commit_at":59,"category_tags":60,"status":17},4721,"markitdown","microsoft\u002Fmarkitdown","MarkItDown 是一款由微软 AutoGen 团队打造的轻量级 Python 工具，专为将各类文件高效转换为 Markdown 格式而设计。它支持 PDF、Word、Excel、PPT、图片（含 OCR）、音频（含语音转录）、HTML 乃至 YouTube 链接等多种格式的解析，能够精准提取文档中的标题、列表、表格和链接等关键结构信息。\n\n在人工智能应用日益普及的今天，大语言模型（LLM）虽擅长处理文本，却难以直接读取复杂的二进制办公文档。MarkItDown 恰好解决了这一痛点，它将非结构化或半结构化的文件转化为模型“原生理解”且 Token 效率极高的 Markdown 格式，成为连接本地文件与 AI 分析 pipeline 的理想桥梁。此外，它还提供了 MCP（模型上下文协议）服务器，可无缝集成到 Claude Desktop 等 LLM 应用中。\n\n这款工具特别适合开发者、数据科学家及 AI 研究人员使用，尤其是那些需要构建文档检索增强生成（RAG）系统、进行批量文本分析或希望让 AI 助手直接“阅读”本地文件的用户。虽然生成的内容也具备一定可读性，但其核心优势在于为机器",93400,"2026-04-06T19:52:38",[52,14],{"id":62,"github_repo":63,"name":64,"description_en":65,"description_zh":66,"ai_summary_zh":66,"readme_en":67,"readme_zh":68,"quickstart_zh":69,"use_case_zh":70,"hero_image_url":71,"owner_login":72,"owner_name":73,"owner_avatar_url":74,"owner_bio":75,"owner_company":76,"owner_location":76,"owner_email":76,"owner_twitter":77,"owner_website":78,"owner_url":79,"languages":80,"stars":97,"forks":98,"last_commit_at":99,"license":100,"difficulty_score":32,"env_os":101,"env_gpu":102,"env_ram":103,"env_deps":104,"category_tags":110,"github_topics":76,"view_count":32,"oss_zip_url":76,"oss_zip_packed_at":76,"status":17,"created_at":111,"updated_at":112,"faqs":113,"releases":143},6985,"BeehiveInnovations\u002Fpal-mcp-server","pal-mcp-server","The power of Claude Code \u002F GeminiCLI \u002F CodexCLI + [Gemini \u002F OpenAI \u002F OpenRouter \u002F Azure \u002F Grok \u002F Ollama \u002F Custom Model \u002F All Of The Above] working as one.","pal-mcp-server 是一款强大的 AI 模型抽象层工具，旨在将您钟爱的命令行界面（如 Claude Code、Gemini CLI、Codex CLI 或 Cursor）与市面上几乎所有主流 AI 模型无缝连接。它打破了单一模型的限制，让您能在同一个工作流中灵活调用 Gemini、OpenAI、Grok、Ollama 等多种模型，仿佛拥有了一支各司其职的 AI 开发团队。\n\n该工具主要解决了开发者在复杂任务中受限于单个模型能力或上下文窗口的问题。通过 pal-mcp-server，您可以让不同模型针对同一问题进行“辩论”以获取更优解，或将代码审查、Bug 排查等重型任务交给独立的子代理处理，从而避免主会话的上下文被污染，保持工作环境整洁高效。\n\n它特别适合需要处理复杂工程任务、追求高质量代码产出以及希望精细化控制 AI 工作流的软件开发者和技术研究人员。其独特的技术亮点在于新增的\"clink\"功能，支持 CLI 间的直接桥接与子代理生成，实现了真正的上下文隔离与角色专业化；同时具备对话线程延续能力，确保多模型协作时的信息流畅传递，让从方案探讨到最终落地的全流程更加顺滑自然。","# PAL MCP: Many Workflows. One Context.\n\n\u003Cdiv align=\"center\">\n\n  \u003Cem>Your AI's PAL – a Provider Abstraction Layer\u003C\u002Fem>\u003Cbr \u002F>\n  \u003Csub>\u003Ca href=\"docs\u002Fname-change.md\">Formerly known as Zen MCP\u003C\u002Fa>\u003C\u002Fsub>\n\n  [PAL in action](https:\u002F\u002Fgithub.com\u002Fuser-attachments\u002Fassets\u002F0d26061e-5f21-4ab1-b7d0-f883ddc2c3da)\n\n👉 **[Watch more examples](#-watch-tools-in-action)**\n\n### Your CLI + Multiple Models = Your AI Dev Team\n\n**Use the 🤖 CLI you love:**  \n[Claude Code](https:\u002F\u002Fwww.anthropic.com\u002Fclaude-code) · [Gemini CLI](https:\u002F\u002Fgithub.com\u002Fgoogle-gemini\u002Fgemini-cli) · [Codex CLI](https:\u002F\u002Fgithub.com\u002Fopenai\u002Fcodex) · [Qwen Code CLI](https:\u002F\u002Fqwenlm.github.io\u002Fqwen-code-docs\u002F) · [Cursor](https:\u002F\u002Fcursor.com) · _and more_\n\n**With multiple models within a single prompt:**  \nGemini · OpenAI · Anthropic · Grok · Azure · Ollama · OpenRouter · DIAL · On-Device Model\n\n\u003C\u002Fdiv>\n\n---\n\n## 🆕 Now with CLI-to-CLI Bridge\n\nThe new **[`clink`](docs\u002Ftools\u002Fclink.md)** (CLI + Link) tool connects external AI CLIs directly into your workflow:\n\n- **Connect external CLIs** like [Gemini CLI](https:\u002F\u002Fgithub.com\u002Fgoogle-gemini\u002Fgemini-cli), [Codex CLI](https:\u002F\u002Fgithub.com\u002Fopenai\u002Fcodex), and [Claude Code](https:\u002F\u002Fwww.anthropic.com\u002Fclaude-code) directly into your workflow\n- **CLI Subagents** - Launch isolated CLI instances from _within_ your current CLI! Claude Code can spawn Codex subagents, Codex can spawn Gemini CLI subagents, etc. Offload heavy tasks (code reviews, bug hunting) to fresh contexts while your main session's context window remains unpolluted. Each subagent returns only final results.\n- **Context Isolation** - Run separate investigations without polluting your primary workspace\n- **Role Specialization** - Spawn `planner`, `codereviewer`, or custom role agents with specialized system prompts\n- **Full CLI Capabilities** - Web search, file inspection, MCP tool access, latest documentation lookups\n- **Seamless Continuity** - Sub-CLIs participate as first-class members with full conversation context between tools\n\n```bash\n# Codex spawns Codex subagent for isolated code review in fresh context\nclink with codex codereviewer to audit auth module for security issues\n# Subagent reviews in isolation, returns final report without cluttering your context as codex reads each file and walks the directory structure\n\n# Consensus from different AI models → Implementation handoff with full context preservation between tools\nUse consensus with gpt-5 and gemini-pro to decide: dark mode or offline support next\nContinue with clink gemini - implement the recommended feature\n# Gemini receives full debate context and starts coding immediately\n```\n\n👉 **[Learn more about clink](docs\u002Ftools\u002Fclink.md)**\n\n---\n\n## Why PAL MCP?\n\n**Why rely on one AI model when you can orchestrate them all?**\n\nA Model Context Protocol server that supercharges tools like [Claude Code](https:\u002F\u002Fwww.anthropic.com\u002Fclaude-code), [Codex CLI](https:\u002F\u002Fdevelopers.openai.com\u002Fcodex\u002Fcli), and IDE clients such\nas [Cursor](https:\u002F\u002Fcursor.com) or the [Claude Dev VS Code extension](https:\u002F\u002Fmarketplace.visualstudio.com\u002Fitems?itemName=Anthropic.claude-vscode). **PAL MCP connects your favorite AI tool\nto multiple AI models** for enhanced code analysis, problem-solving, and collaborative development.\n\n### True AI Collaboration with Conversation Continuity\n\nPAL supports **conversation threading** so your CLI can **discuss ideas with multiple AI models, exchange reasoning, get second opinions, and even run collaborative debates between models** to help you reach deeper insights and better solutions.\n\nYour CLI always stays in control but gets perspectives from the best AI for each subtask. Context carries forward seamlessly across tools and models, enabling complex workflows like: code reviews with multiple models → automated planning → implementation → pre-commit validation.\n\n> **You're in control.** Your CLI of choice orchestrates the AI team, but you decide the workflow. Craft powerful prompts that bring in Gemini Pro, GPT 5, Flash, or local offline models exactly when needed.\n\n\u003Cdetails>\n\u003Csummary>\u003Cb>Reasons to Use PAL MCP\u003C\u002Fb>\u003C\u002Fsummary>\n\nA typical workflow with Claude Code as an example:\n\n1. **Multi-Model Orchestration** - Claude coordinates with Gemini Pro, O3, GPT-5, and 50+ other models to get the best analysis for each task\n\n2. **Context Revival Magic** - Even after Claude's context resets, continue conversations seamlessly by having other models \"remind\" Claude of the discussion\n\n3. **Guided Workflows** - Enforces systematic investigation phases that prevent rushed analysis and ensure thorough code examination\n\n4. **Extended Context Windows** - Break Claude's limits by delegating to Gemini (1M tokens) or O3 (200K tokens) for massive codebases\n\n5. **True Conversation Continuity** - Full context flows across tools and models - Gemini remembers what O3 said 10 steps ago\n\n6. **Model-Specific Strengths** - Extended thinking with Gemini Pro, blazing speed with Flash, strong reasoning with O3, privacy with local Ollama\n\n7. **Professional Code Reviews** - Multi-pass analysis with severity levels, actionable feedback, and consensus from multiple AI experts\n\n8. **Smart Debugging Assistant** - Systematic root cause analysis with hypothesis tracking and confidence levels\n\n9. **Automatic Model Selection** - Claude intelligently picks the right model for each subtask (or you can specify)\n\n10. **Vision Capabilities** - Analyze screenshots, diagrams, and visual content with vision-enabled models\n\n11. **Local Model Support** - Run Llama, Mistral, or other models locally for complete privacy and zero API costs\n\n12. **Bypass MCP Token Limits** - Automatically works around MCP's 25K limit for large prompts and responses\n\n**The Killer Feature:** When Claude's context resets, just ask to \"continue with O3\" - the other model's response magically revives Claude's understanding without re-ingesting documents!\n\n#### Example: Multi-Model Code Review Workflow\n\n1. `Perform a codereview using gemini pro and o3 and use planner to generate a detailed plan, implement the fixes and do a final precommit check by continuing from the previous codereview`\n2. This triggers a [`codereview`](docs\u002Ftools\u002Fcodereview.md) workflow where Claude walks the code, looking for all kinds of issues\n3. After multiple passes, collects relevant code and makes note of issues along the way\n4. Maintains a `confidence` level between `exploring`, `low`, `medium`, `high` and `certain` to track how confidently it's been able to find and identify issues\n5. Generates a detailed list of critical -> low issues\n6. Shares the relevant files, findings, etc with **Gemini Pro** to perform a deep dive for a second [`codereview`](docs\u002Ftools\u002Fcodereview.md)\n7. Comes back with a response and next does the same with o3, adding to the prompt if a new discovery comes to light\n8. When done, Claude takes in all the feedback and combines a single list of all critical -> low issues, including good patterns in your code. The final list includes new findings or revisions in case Claude misunderstood or missed something crucial and one of the other models pointed this out\n9. It then uses the [`planner`](docs\u002Ftools\u002Fplanner.md) workflow to break the work down into simpler steps if a major refactor is required\n10. Claude then performs the actual work of fixing highlighted issues\n11. When done, Claude returns to Gemini Pro for a [`precommit`](docs\u002Ftools\u002Fprecommit.md) review\n\nAll within a single conversation thread! Gemini Pro in step 11 _knows_ what was recommended by O3 in step 7! Taking that context\nand review into consideration to aid with its final pre-commit review.\n\n**Think of it as Claude Code _for_ Claude Code.** This MCP isn't magic. It's just **super-glue**.\n\n> **Remember:** Claude stays in full control — but **YOU** call the shots.\n> PAL is designed to have Claude engage other models only when needed — and to follow through with meaningful back-and-forth.\n> **You're** the one who crafts the powerful prompt that makes Claude bring in Gemini, Flash, O3 — or fly solo.\n> You're the guide. The prompter. The puppeteer.\n> #### You are the AI - **Actually Intelligent**.\n\u003C\u002Fdetails>\n\n#### Recommended AI Stack\n\n\u003Cdetails>\n\u003Csummary>For Claude Code Users\u003C\u002Fsummary>\n\nFor best results when using [Claude Code](https:\u002F\u002Fclaude.ai\u002Fcode):  \n\n- **Sonnet 4.5** - All agentic work and orchestration\n- **Gemini 3.0 Pro** OR **GPT-5.2 \u002F Pro** - Deep thinking, additional code reviews, debugging and validations, pre-commit analysis\n\u003C\u002Fdetails>\n\n\u003Cdetails>\n\u003Csummary>For Codex Users\u003C\u002Fsummary>\n\nFor best results when using [Codex CLI](https:\u002F\u002Fdevelopers.openai.com\u002Fcodex\u002Fcli):  \n\n- **GPT-5.2 Codex Medium** - All agentic work and orchestration\n- **Gemini 3.0 Pro** OR **GPT-5.2-Pro** - Deep thinking, additional code reviews, debugging and validations, pre-commit analysis\n\u003C\u002Fdetails>\n\n## Quick Start (5 minutes)\n\n**Prerequisites:** Python 3.10+, Git, [uv installed](https:\u002F\u002Fdocs.astral.sh\u002Fuv\u002Fgetting-started\u002Finstallation\u002F)\n\n**1. Get API Keys** (choose one or more):\n- **[OpenRouter](https:\u002F\u002Fopenrouter.ai\u002F)** - Access multiple models with one API\n- **[Gemini](https:\u002F\u002Fmakersuite.google.com\u002Fapp\u002Fapikey)** - Google's latest models\n- **[OpenAI](https:\u002F\u002Fplatform.openai.com\u002Fapi-keys)** - O3, GPT-5 series\n- **[Azure OpenAI](https:\u002F\u002Flearn.microsoft.com\u002Fazure\u002Fai-services\u002Fopenai\u002F)** - Enterprise deployments of GPT-4o, GPT-4.1, GPT-5 family\n- **[X.AI](https:\u002F\u002Fconsole.x.ai\u002F)** - Grok models\n- **[DIAL](https:\u002F\u002Fdialx.ai\u002F)** - Vendor-agnostic model access\n- **[Ollama](https:\u002F\u002Follama.ai\u002F)** - Local models (free)\n\n**2. Install** (choose one):\n\n**Option A: Clone and Automatic Setup** (recommended)\n```bash\ngit clone https:\u002F\u002Fgithub.com\u002FBeehiveInnovations\u002Fpal-mcp-server.git\ncd pal-mcp-server\n\n# Handles everything: setup, config, API keys from system environment. \n# Auto-configures Claude Desktop, Claude Code, Gemini CLI, Codex CLI, Qwen CLI\n# Enable \u002F disable additional settings in .env\n.\u002Frun-server.sh  \n```\n\n**Option B: Instant Setup with [uvx](https:\u002F\u002Fdocs.astral.sh\u002Fuv\u002Fgetting-started\u002Finstallation\u002F)**\n```json\n\u002F\u002F Add to ~\u002F.claude\u002Fsettings.json or .mcp.json\n\u002F\u002F Don't forget to add your API keys under env\n{\n  \"mcpServers\": {\n    \"pal\": {\n      \"command\": \"bash\",\n      \"args\": [\"-c\", \"for p in $(which uvx 2>\u002Fdev\u002Fnull) $HOME\u002F.local\u002Fbin\u002Fuvx \u002Fopt\u002Fhomebrew\u002Fbin\u002Fuvx \u002Fusr\u002Flocal\u002Fbin\u002Fuvx uvx; do [ -x \\\"$p\\\" ] && exec \\\"$p\\\" --from git+https:\u002F\u002Fgithub.com\u002FBeehiveInnovations\u002Fpal-mcp-server.git pal-mcp-server; done; echo 'uvx not found' >&2; exit 1\"],\n      \"env\": {\n        \"PATH\": \"\u002Fusr\u002Flocal\u002Fbin:\u002Fusr\u002Fbin:\u002Fbin:\u002Fopt\u002Fhomebrew\u002Fbin:~\u002F.local\u002Fbin\",\n        \"GEMINI_API_KEY\": \"your-key-here\",\n        \"DISABLED_TOOLS\": \"analyze,refactor,testgen,secaudit,docgen,tracer\",\n        \"DEFAULT_MODEL\": \"auto\"\n      }\n    }\n  }\n}\n```\n\n**3. Start Using!**\n```\n\"Use pal to analyze this code for security issues with gemini pro\"\n\"Debug this error with o3 and then get flash to suggest optimizations\"\n\"Plan the migration strategy with pal, get consensus from multiple models\"\n\"clink with cli_name=\\\"gemini\\\" role=\\\"planner\\\" to draft a phased rollout plan\"\n```\n\n👉 **[Complete Setup Guide](docs\u002Fgetting-started.md)** with detailed installation, configuration for Gemini \u002F Codex \u002F Qwen, and troubleshooting\n👉 **[Cursor & VS Code Setup](docs\u002Fgetting-started.md#ide-clients)** for IDE integration instructions\n📺 **[Watch tools in action](#-watch-tools-in-action)** to see real-world examples\n\n## Provider Configuration\n\nPAL activates any provider that has credentials in your `.env`. See `.env.example` for deeper customization.\n\n## Core Tools\n\n> **Note:** Each tool comes with its own multi-step workflow, parameters, and descriptions that consume valuable context window space even when not in use. To optimize performance, some tools are disabled by default. See [Tool Configuration](#tool-configuration) below to enable them.\n\n**Collaboration & Planning** *(Enabled by default)*\n- **[`clink`](docs\u002Ftools\u002Fclink.md)** - Bridge requests to external AI CLIs (Gemini planner, codereviewer, etc.)\n- **[`chat`](docs\u002Ftools\u002Fchat.md)** - Brainstorm ideas, get second opinions, validate approaches. With capable models (GPT-5.2 Pro, Gemini 3.0 Pro), generates complete code \u002F implementation\n- **[`thinkdeep`](docs\u002Ftools\u002Fthinkdeep.md)** - Extended reasoning, edge case analysis, alternative perspectives\n- **[`planner`](docs\u002Ftools\u002Fplanner.md)** - Break down complex projects into structured, actionable plans\n- **[`consensus`](docs\u002Ftools\u002Fconsensus.md)** - Get expert opinions from multiple AI models with stance steering\n\n**Code Analysis & Quality**\n- **[`debug`](docs\u002Ftools\u002Fdebug.md)** - Systematic investigation and root cause analysis\n- **[`precommit`](docs\u002Ftools\u002Fprecommit.md)** - Validate changes before committing, prevent regressions\n- **[`codereview`](docs\u002Ftools\u002Fcodereview.md)** - Professional reviews with severity levels and actionable feedback\n- **[`analyze`](docs\u002Ftools\u002Fanalyze.md)** *(disabled by default - [enable](#tool-configuration))* - Understand architecture, patterns, dependencies across entire codebases\n\n**Development Tools** *(Disabled by default - [enable](#tool-configuration))*\n- **[`refactor`](docs\u002Ftools\u002Frefactor.md)** - Intelligent code refactoring with decomposition focus\n- **[`testgen`](docs\u002Ftools\u002Ftestgen.md)** - Comprehensive test generation with edge cases\n- **[`secaudit`](docs\u002Ftools\u002Fsecaudit.md)** - Security audits with OWASP Top 10 analysis\n- **[`docgen`](docs\u002Ftools\u002Fdocgen.md)** - Generate documentation with complexity analysis\n\n**Utilities**\n- **[`apilookup`](docs\u002Ftools\u002Fapilookup.md)** - Forces current-year API\u002FSDK documentation lookups in a sub-process (saves tokens within the current context window), prevents outdated training data responses\n- **[`challenge`](docs\u002Ftools\u002Fchallenge.md)** - Prevent \"You're absolutely right!\" responses with critical analysis\n- **[`tracer`](docs\u002Ftools\u002Ftracer.md)** *(disabled by default - [enable](#tool-configuration))* - Static analysis prompts for call-flow mapping\n\n\u003Cdetails>\n\u003Csummary>\u003Cb id=\"tool-configuration\">👉 Tool Configuration\u003C\u002Fb>\u003C\u002Fsummary>\n\n### Default Configuration\n\nTo optimize context window usage, only essential tools are enabled by default:\n\n**Enabled by default:**\n- `chat`, `thinkdeep`, `planner`, `consensus` - Core collaboration tools\n- `codereview`, `precommit`, `debug` - Essential code quality tools\n- `apilookup` - Rapid API\u002FSDK information lookup\n- `challenge` - Critical thinking utility\n\n**Disabled by default:**\n- `analyze`, `refactor`, `testgen`, `secaudit`, `docgen`, `tracer`\n\n### Enabling Additional Tools\n\nTo enable additional tools, remove them from the `DISABLED_TOOLS` list:\n\n**Option 1: Edit your .env file**\n```bash\n# Default configuration (from .env.example)\nDISABLED_TOOLS=analyze,refactor,testgen,secaudit,docgen,tracer\n\n# To enable specific tools, remove them from the list\n# Example: Enable analyze tool\nDISABLED_TOOLS=refactor,testgen,secaudit,docgen,tracer\n\n# To enable ALL tools\nDISABLED_TOOLS=\n```\n\n**Option 2: Configure in MCP settings**\n```json\n\u002F\u002F In ~\u002F.claude\u002Fsettings.json or .mcp.json\n{\n  \"mcpServers\": {\n    \"pal\": {\n      \"env\": {\n        \u002F\u002F Tool configuration\n        \"DISABLED_TOOLS\": \"refactor,testgen,secaudit,docgen,tracer\",\n        \"DEFAULT_MODEL\": \"pro\",\n        \"DEFAULT_THINKING_MODE_THINKDEEP\": \"high\",\n        \n        \u002F\u002F API configuration\n        \"GEMINI_API_KEY\": \"your-gemini-key\",\n        \"OPENAI_API_KEY\": \"your-openai-key\",\n        \"OPENROUTER_API_KEY\": \"your-openrouter-key\",\n        \n        \u002F\u002F Logging and performance\n        \"LOG_LEVEL\": \"INFO\",\n        \"CONVERSATION_TIMEOUT_HOURS\": \"6\",\n        \"MAX_CONVERSATION_TURNS\": \"50\"\n      }\n    }\n  }\n}\n```\n\n**Option 3: Enable all tools**\n```json\n\u002F\u002F Remove or empty the DISABLED_TOOLS to enable everything\n{\n  \"mcpServers\": {\n    \"pal\": {\n      \"env\": {\n        \"DISABLED_TOOLS\": \"\"\n      }\n    }\n  }\n}\n```\n\n**Note:**\n- Essential tools (`version`, `listmodels`) cannot be disabled\n- After changing tool configuration, restart your Claude session for changes to take effect\n- Each tool adds to context window usage, so only enable what you need\n\n\u003C\u002Fdetails>\n\n## 📺 Watch Tools In Action\n\n\u003Cdetails>\n\u003Csummary>\u003Cb>Chat Tool\u003C\u002Fb> - Collaborative decision making and multi-turn conversations\u003C\u002Fsummary>\n\n**Picking Redis vs Memcached:**\n\n[Chat Redis or Memcached_web.webm](https:\u002F\u002Fgithub.com\u002Fuser-attachments\u002Fassets\u002F41076cfe-dd49-4dfc-82f5-d7461b34705d)\n\n**Multi-turn conversation with continuation:**\n\n[Chat With Gemini_web.webm](https:\u002F\u002Fgithub.com\u002Fuser-attachments\u002Fassets\u002F37bd57ca-e8a6-42f7-b5fb-11de271e95db)\n\n\u003C\u002Fdetails>\n\n\u003Cdetails>\n\u003Csummary>\u003Cb>Consensus Tool\u003C\u002Fb> - Multi-model debate and decision making\u003C\u002Fsummary>\n\n**Multi-model consensus debate:**\n\n[PAL Consensus Debate](https:\u002F\u002Fgithub.com\u002Fuser-attachments\u002Fassets\u002F76a23dd5-887a-4382-9cf0-642f5cf6219e)\n\n\u003C\u002Fdetails>\n\n\u003Cdetails>\n\u003Csummary>\u003Cb>PreCommit Tool\u003C\u002Fb> - Comprehensive change validation\u003C\u002Fsummary>\n\n**Pre-commit validation workflow:**\n\n\u003Cdiv align=\"center\">\n  \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002FBeehiveInnovations_pal-mcp-server_readme_045584c73c01.png\" width=\"950\">\n\u003C\u002Fdiv>\n\n\u003C\u002Fdetails>\n\n\u003Cdetails>\n\u003Csummary>\u003Cb>API Lookup Tool\u003C\u002Fb> - Current vs outdated API documentation\u003C\u002Fsummary>\n\n**Without PAL - outdated APIs:**\n\n[API without PAL](https:\u002F\u002Fgithub.com\u002Fuser-attachments\u002Fassets\u002F01a79dc9-ad16-4264-9ce1-76a56c3580ee)\n\n**With PAL - current APIs:**\n\n[API with PAL](https:\u002F\u002Fgithub.com\u002Fuser-attachments\u002Fassets\u002F5c847326-4b66-41f7-8f30-f380453dce22)\n\n\u003C\u002Fdetails>\n\n\u003Cdetails>\n\u003Csummary>\u003Cb>Challenge Tool\u003C\u002Fb> - Critical thinking vs reflexive agreement\u003C\u002Fsummary>\n\n**Without PAL:**\n\n![without_pal@2x](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002FBeehiveInnovations_pal-mcp-server_readme_c19c869c8ff6.png)\n\n**With PAL:**\n\n![with_pal@2x](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002FBeehiveInnovations_pal-mcp-server_readme_855b6613d596.png)\n\n\u003C\u002Fdetails>\n\n## Key Features\n\n**AI Orchestration**\n- **Auto model selection** - Claude picks the right AI for each task\n- **Multi-model workflows** - Chain different models in single conversations\n- **Conversation continuity** - Context preserved across tools and models\n- **[Context revival](docs\u002Fcontext-revival.md)** - Continue conversations even after context resets\n\n**Model Support**\n- **Multiple providers** - Gemini, OpenAI, Azure, X.AI, OpenRouter, DIAL, Ollama\n- **Latest models** - GPT-5, Gemini 3.0 Pro, O3, Grok-4, local Llama\n- **[Thinking modes](docs\u002Fadvanced-usage.md#thinking-modes)** - Control reasoning depth vs cost\n- **Vision support** - Analyze images, diagrams, screenshots\n\n**Developer Experience**\n- **Guided workflows** - Systematic investigation prevents rushed analysis\n- **Smart file handling** - Auto-expand directories, manage token limits\n- **Web search integration** - Access current documentation and best practices\n- **[Large prompt support](docs\u002Fadvanced-usage.md#working-with-large-prompts)** - Bypass MCP's 25K token limit\n\n## Example Workflows\n\n**Multi-model Code Review:**\n```\n\"Perform a codereview using gemini pro and o3, then use planner to create a fix strategy\"\n```\n→ Claude reviews code systematically → Consults Gemini Pro → Gets O3's perspective → Creates unified action plan\n\n**Collaborative Debugging:**\n```\n\"Debug this race condition with max thinking mode, then validate the fix with precommit\"\n```\n→ Deep investigation → Expert analysis → Solution implementation → Pre-commit validation\n\n**Architecture Planning:**\n```\n\"Plan our microservices migration, get consensus from pro and o3 on the approach\"\n```\n→ Structured planning → Multiple expert opinions → Consensus building → Implementation roadmap\n\n👉 **[Advanced Usage Guide](docs\u002Fadvanced-usage.md)** for complex workflows, model configuration, and power-user features\n\n## Quick Links\n\n**📖 Documentation**\n- [Docs Overview](docs\u002Findex.md) - High-level map of major guides\n- [Getting Started](docs\u002Fgetting-started.md) - Complete setup guide\n- [Tools Reference](docs\u002Ftools\u002F) - All tools with examples\n- [Advanced Usage](docs\u002Fadvanced-usage.md) - Power user features\n- [Configuration](docs\u002Fconfiguration.md) - Environment variables, restrictions\n- [Adding Providers](docs\u002Fadding_providers.md) - Provider-specific setup (OpenAI, Azure, custom gateways)\n- [Model Ranking Guide](docs\u002Fmodel_ranking.md) - How intelligence scores drive auto-mode suggestions\n\n**🔧 Setup & Support**\n- [WSL Setup](docs\u002Fwsl-setup.md) - Windows users\n- [Troubleshooting](docs\u002Ftroubleshooting.md) - Common issues\n- [Contributing](docs\u002Fcontributions.md) - Code standards, PR process\n\n## License\n\nApache 2.0 License - see [LICENSE](LICENSE) file for details.\n\n## Acknowledgments\n\nBuilt with the power of **Multi-Model AI** collaboration 🤝\n- **A**ctual **I**ntelligence by real Humans\n- [MCP (Model Context Protocol)](https:\u002F\u002Fmodelcontextprotocol.com)\n- [Codex CLI](https:\u002F\u002Fdevelopers.openai.com\u002Fcodex\u002Fcli)\n- [Claude Code](https:\u002F\u002Fclaude.ai\u002Fcode)\n- [Gemini](https:\u002F\u002Fai.google.dev\u002F)\n- [OpenAI](https:\u002F\u002Fopenai.com\u002F)\n- [Azure OpenAI](https:\u002F\u002Flearn.microsoft.com\u002Fazure\u002Fai-services\u002Fopenai\u002F)\n\n### Star History\n\n[![Star History Chart](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002FBeehiveInnovations_pal-mcp-server_readme_a41d3b421544.png)](https:\u002F\u002Fwww.star-history.com\u002F#BeehiveInnovations\u002Fpal-mcp-server&Date)\n","# PAL MCP：多种工作流，一个上下文。\n\n\u003Cdiv align=\"center\">\n\n  \u003Cem>你的 AI 的 PAL——提供商抽象层\u003C\u002Fem>\u003Cbr \u002F>\n  \u003Csub>\u003Ca href=\"docs\u002Fname-change.md\">曾用名：Zen MCP\u003C\u002Fa>\u003C\u002Fsub>\n\n  [PAL 实战](https:\u002F\u002Fgithub.com\u002Fuser-attachments\u002Fassets\u002F0d26061e-5f21-4ab1-b7d0-f883ddc2c3da)\n\n👉 **[观看更多示例](#-watch-tools-in-action)**\n\n### 你的命令行工具 + 多种模型 = 你的 AI 开发团队\n\n**使用你喜爱的 🤖 命令行工具：**  \n[Claude Code](https:\u002F\u002Fwww.anthropic.com\u002Fclaude-code) · [Gemini CLI](https:\u002F\u002Fgithub.com\u002Fgoogle-gemini\u002Fgemini-cli) · [Codex CLI](https:\u002F\u002Fgithub.com\u002Fopenai\u002Fcodex) · [Qwen Code CLI](https:\u002F\u002Fqwenlm.github.io\u002Fqwen-code-docs\u002F) · [Cursor](https:\u002F\u002Fcursor.com) · _以及其他_\n\n**在一个提示中集成多种模型：**  \nGemini · OpenAI · Anthropic · Grok · Azure · Ollama · OpenRouter · DIAL · 设备端模型\n\n\u003C\u002Fdiv>\n\n---\n\n## 🆕 现已支持 CLI 到 CLI 桥接\n\n全新的 **[`clink`](docs\u002Ftools\u002Fclink.md)**（CLI + 链接）工具可将外部 AI 命令行工具直接接入你的工作流：\n\n- **连接外部 CLI**，如 [Gemini CLI](https:\u002F\u002Fgithub.com\u002Fgoogle-gemini\u002Fgemini-cli)、[Codex CLI](https:\u002F\u002Fgithub.com\u002Fopenai\u002Fcodex) 和 [Claude Code](https:\u002F\u002Fwww.anthropic.com\u002Fclaude-code)，直接融入你的工作流程\n- **CLI 子代理**——从当前 CLI 内部启动独立的 CLI 实例！Claude Code 可以派生 Codex 子代理，Codex 可以派生 Gemini CLI 子代理，等等。将繁重的任务（代码审查、缺陷排查）交由全新上下文中执行，同时保持主会话的上下文窗口不受干扰。每个子代理仅返回最终结果。\n- **上下文隔离**——在不污染主工作区的情况下开展独立调查\n- **角色专业化**——根据特定系统提示，派生 `planner`、`codereviewer` 或自定义角色代理\n- **完整 CLI 功能**——网络搜索、文件检查、MCP 工具访问、最新文档查询\n- **无缝衔接**——子 CLI 作为一级成员参与，各工具间保持完整的对话上下文\n\n```bash\n# Codex 启动 Codex 子代理，在全新上下文中进行隔离式代码审查\nclink with codex codereviewer to audit auth module for security issues\n# 子代理在隔离环境中完成审查，仅返回最终报告，不会因 Codex 逐个读取文件和遍历目录结构而使你的上下文混乱\n\n# 不同 AI 模型达成共识 → 在完全保留工具间上下文的情况下移交实现任务\n利用 GPT-5 和 Gemini Pro 的共识来决定：接下来是开发暗黑模式还是离线支持\n继续使用 clink gemini 来实现推荐的功能\n# Gemini 接收到完整的讨论上下文，并立即开始编码\n```\n\n👉 **[了解更多关于 clink 的信息](docs\u002Ftools\u002Fclink.md)**\n\n---\n\n## 为什么选择 PAL MCP？\n\n**为何只依赖单一 AI 模型，而不将其全部编排起来呢？**\n\nPAL MCP 是一种模型上下文协议服务器，能够为 [Claude Code](https:\u002F\u002Fwww.anthropic.com\u002Fclaude-code)、[Codex CLI](https:\u002F\u002Fdevelopers.openai.com\u002Fcodex\u002Fcli) 等工具以及 [Cursor](https:\u002F\u002Fcursor.com) 或 [Claude Dev VS Code 扩展](https:\u002F\u002Fmarketplace.visualstudio.com\u002Fitems?itemName=Anthropic.claude-vscode) 等 IDE 客户端提供强大助力。**PAL MCP 将你最喜爱的 AI 工具与多种 AI 模型连接起来**，从而提升代码分析、问题解决和协作开发的能力。\n\n### 真正的 AI 协作与对话连续性\n\nPAL 支持 **对话线程化**，因此您的 CLI 可以 **与多个 AI 模型讨论想法、交换推理过程、获取第二意见，甚至在模型之间进行协作式辩论**，从而帮助您获得更深入的洞察和更好的解决方案。\n\n您的 CLI 始终掌握主动权，但会在每个子任务中从最适合的 AI 获取视角。上下文可以在不同工具和模型之间无缝传递，从而实现复杂的流程，例如：使用多个模型进行代码审查 → 自动化规划 → 实施 → 提交前验证。\n\n> **您掌控一切。** 您选择的 CLI 将协调整个 AI 团队，而工作流则由您决定。您可以编写强大的提示词，在需要时精准调用 Gemini Pro、GPT 5、Flash 或本地离线模型。\n\n\u003Cdetails>\n\u003Csummary>\u003Cb>使用 PAL MCP 的理由\u003C\u002Fb>\u003C\u002Fsummary>\n\n以 Claude Code 为例的典型工作流程：\n\n1. **多模型编排** - Claude 会与 Gemini Pro、O3、GPT-5 以及其他 50 多个模型协同工作，为每个任务提供最佳分析。\n\n2. **上下文恢复魔法** - 即使 Claude 的上下文重置，也可以通过其他模型“提醒”Claude 讨论内容，从而无缝延续对话。\n\n3. **引导式工作流程** - 强制执行系统化的调查阶段，防止仓促分析，并确保对代码进行全面检查。\n\n4. **扩展的上下文窗口** - 通过将任务委托给 Gemini（100 万 token）或 O3（20 万 token），突破 Claude 的限制，处理超大规模代码库。\n\n5. **真正的对话连续性** - 完整的上下文可在不同工具和模型之间流动——Gemini 能够记住 O3 在 10 步之前说过的内容。\n\n6. **模型特有优势** - 使用 Gemini Pro 进行深度思考，借助 Flash 实现极速响应，依靠 O3 进行强逻辑推理，同时利用本地 Ollama 保障隐私。\n\n7. **专业的代码审查** - 采用多轮分析，按严重程度分级，提供可操作的反馈，并由多位 AI 专家达成共识。\n\n8. **智能调试助手** - 通过跟踪假设和置信度，系统地分析根本原因。\n\n9. **自动模型选择** - Claude 会智能地为每个子任务选择合适的模型（当然，您也可以手动指定）。\n\n10. **视觉能力** - 利用具备视觉功能的模型分析截图、图表和其他视觉内容。\n\n11. **本地模型支持** - 在本地运行 Llama、Mistral 等模型，以实现完全隐私且无需支付 API 费用。\n\n12. **绕过 MCP 的 token 限制** - 自动解决 MCP 对大型提示和响应的 2.5 万 token 限制问题。\n\n**杀手级功能：** 当 Claude 的上下文重置时，只需请求“继续与 O3 对话”，另一模型的回复就会神奇地恢复 Claude 的理解，而无需重新输入文档！\n\n#### 示例：多模型代码审查工作流程\n\n1. `使用 Gemini Pro 和 O3 进行代码审查，并利用规划器生成详细计划，实施修复后，继续之前的代码审查进行最终的提交前检查`\n2. 这将触发一个 [`codereview`](docs\u002Ftools\u002Fcodereview.md) 工作流，Claude 会逐行检查代码，寻找各类问题。\n3. 经过多轮检查，收集相关代码并记录发现的问题。\n4. 维持一个 `confidence` 级别，范围从 `exploring`、`low`、`medium`、`high` 到 `certain`，以跟踪其发现和识别问题的自信程度。\n5. 生成一份详细的严重性从高到低的问题清单。\n6. 将相关文件和发现等内容分享给 **Gemini Pro**，以进行第二次深入的 [`codereview`](docs\u002Ftools\u002Fcodereview.md)。\n7. Gemini Pro 返回结果后，Claude 再次与 O3 进行同样的流程，如果过程中有新发现，则在提示中补充相关信息。\n8. 完成后，Claude 整合所有反馈，合并成一份涵盖严重到轻微问题的统一清单，其中还包括代码中的良好实践。最终清单会包含新的发现或修正，以防 Claude 曾经误解或遗漏了某些关键点，而其他模型指出了这些问题。\n9. 如果需要进行重大重构，Claude 会使用 [`planner`](docs\u002Ftools\u002Fplanner.md) 工作流将任务分解为更简单的步骤。\n10. 接着，Claude 开始实际修复已标记的问题。\n11. 修复完成后，Claude 再次回到 Gemini Pro 进行 [`precommit`](docs\u002Ftools\u002Fprecommit.md) 审查。\n\n这一切都发生在同一个对话线程中！第 11 步中的 Gemini Pro _知道_ 第 7 步中 O3 的建议！它会结合这些上下文和审查意见，为其最终的提交前审查提供帮助。\n\n**可以把它想象成专门为 Claude Code 设计的 Claude Code。** 这种 MCP 并非魔法，而是一道 **超级胶水**。\n\n> **请记住：** Claude 仍然完全掌控全局——但真正拍板的是 **您**。\n> PAL 的设计宗旨是让 Claude 仅在必要时与其他模型互动，并保持有意义的来回交流。\n> **您** 才是那个编写强大提示词的人，促使 Claude 调用 Gemini、Flash、O3——或者独自完成任务。\n> 您是引导者、提示词的设计者、幕后操纵者。\n> #### 您就是 AI——**真正智能的 AI**。\n\u003C\u002Fdetails>\n\n#### 推荐的 AI 技术栈\n\n\u003Cdetails>\n\u003Csummary>针对 Claude Code 用户\u003C\u002Fsummary>\n\n为了在使用 [Claude Code](https:\u002F\u002Fclaude.ai\u002Fcode) 时获得最佳效果：\n\n- **Sonnet 4.5** - 用于所有代理性和编排工作\n- **Gemini 3.0 Pro** 或 **GPT-5.2 \u002F Pro** - 用于深度思考、额外的代码审查、调试与验证以及提交前分析\n\u003C\u002Fdetails>\n\n\u003Cdetails>\n\u003Csummary>针对 Codex 用户\u003C\u002Fsummary>\n\n为了在使用 [Codex CLI](https:\u002F\u002Fdevelopers.openai.com\u002Fcodex\u002Fcli) 时获得最佳效果：\n\n- **GPT-5.2 Codex Medium** - 用于所有代理性和编排工作\n- **Gemini 3.0 Pro** 或 **GPT-5.2-Pro** - 用于深度思考、额外的代码审查、调试与验证以及提交前分析\n\u003C\u002Fdetails>\n\n## 快速入门（5 分钟）\n\n**先决条件：** Python 3.10+、Git、已安装 [uv](https:\u002F\u002Fdocs.astral.sh\u002Fuv\u002Fgetting-started\u002Finstallation\u002F)\n\n**1. 获取 API 密钥**（选择一个或多个）：\n- **[OpenRouter](https:\u002F\u002Fopenrouter.ai\u002F)** - 通过一个 API 访问多个模型\n- **[Gemini](https:\u002F\u002Fmakersuite.google.com\u002Fapp\u002Fapikey)** - Google 最新推出的模型\n- **[OpenAI](https:\u002F\u002Fplatform.openai.com\u002Fapi-keys)** - O3、GPT-5 系列\n- **[Azure OpenAI](https:\u002F\u002Flearn.microsoft.com\u002Fazure\u002Fai-services\u002Fopenai\u002F)** - GPT-4o、GPT-4.1、GPT-5 系列的企业级部署\n- **[X.AI](https:\u002F\u002Fconsole.x.ai\u002F)** - Grok 模型\n- **[DIAL](https:\u002F\u002Fdialx.ai\u002F)** - 不受供应商限制的模型访问\n- **[Ollama](https:\u002F\u002Follama.ai\u002F)** - 本地模型（免费）\n\n**2. 安装**（选择一种）：\n\n**选项 A：克隆并自动设置**（推荐）\n```bash\ngit clone https:\u002F\u002Fgithub.com\u002FBeehiveInnovations\u002Fpal-mcp-server.git\ncd pal-mcp-server\n\n# 自动完成所有操作：设置、配置、从系统环境加载 API 密钥。\n# 自动配置 Claude Desktop、Claude Code、Gemini CLI、Codex CLI、Qwen CLI\n\n# 在 .env 中启用或禁用附加设置\n.\u002Frun-server.sh  \n```\n\n**选项 B：使用 [uvx](https:\u002F\u002Fdocs.astral.sh\u002Fuv\u002Fgetting-started\u002Finstallation\u002F) 进行快速设置**\n```json\n\u002F\u002F 添加到 ~\u002F.claude\u002Fsettings.json 或 .mcp.json\n\u002F\u002F 别忘了在 env 下添加你的 API 密钥\n{\n  \"mcpServers\": {\n    \"pal\": {\n      \"command\": \"bash\",\n      \"args\": [\"-c\", \"for p in $(which uvx 2>\u002Fdev\u002Fnull) $HOME\u002F.local\u002Fbin\u002Fuvx \u002Fopt\u002Fhomebrew\u002Fbin\u002Fuvx \u002Fusr\u002Flocal\u002Fbin\u002Fuvx uvx; do [ -x \\\"$p\\\" ] && exec \\\"$p\\\" --from git+https:\u002F\u002Fgithub.com\u002FBeehiveInnovations\u002Fpal-mcp-server.git pal-mcp-server; done; echo 'uvx not found' >&2; exit 1\"],\n      \"env\": {\n        \"PATH\": \"\u002Fusr\u002Flocal\u002Fbin:\u002Fusr\u002Fbin:\u002Fbin:\u002Fopt\u002Fhomebrew\u002Fbin:~\u002F.local\u002Fbin\",\n        \"GEMINI_API_KEY\": \"your-key-here\",\n        \"DISABLED_TOOLS\": \"analyze,refactor,testgen,secaudit,docgen,tracer\",\n        \"DEFAULT_MODEL\": \"auto\"\n      }\n    }\n  }\n}\n```\n\n**3. 开始使用！**\n```\n\"使用 pal 用 gemini pro 分析这段代码的安全问题\"\n\"用 o3 调试这个错误，然后让 flash 提供建议进行优化\"\n\"用 pal 制定迁移策略，并从多个模型中获得共识\"\n\"clink with cli_name=\\\"gemini\\\" role=\\\"planner\\\" 来起草分阶段的部署计划\"\n```\n\n👉 **[完整设置指南](docs\u002Fgetting-started.md)**，包含详细的安装、Gemini\u002FCodex\u002FQwen 的配置以及故障排除\n👉 **[Cursor 和 VS Code 设置](docs\u002Fgetting-started.md#ide-clients)**，提供 IDE 集成说明\n📺 **[观看工具的实际应用](#-watch-tools-in-action)**，了解真实场景中的示例\n\n## 提供商配置\n\nPAL 会激活任何在你的 `.env` 文件中有凭据的提供商。请参阅 `.env.example` 以进行更深入的自定义。\n\n## 核心工具\n\n> **注意：** 每个工具都有自己的多步骤工作流程、参数和描述，即使不使用也会占用宝贵的上下文窗口空间。为了优化性能，部分工具默认被禁用。请参阅下方的 [工具配置](#tool-configuration) 以启用它们。\n\n**协作与规划** *(默认启用)*\n- **[`clink`](docs\u002Ftools\u002Fclink.md)** - 将请求桥接到外部 AI CLI（Gemini 规划器、代码审查等）\n- **[`chat`](docs\u002Ftools\u002Fchat.md)** - 头脑风暴想法、获取第二意见、验证方法。在功能强大的模型（如 GPT-5.2 Pro、Gemini 3.0 Pro）下，可生成完整的代码\u002F实现\n- **[`thinkdeep`](docs\u002Ftools\u002Fthinkdeep.md)** - 扩展推理、边缘情况分析、替代视角\n- **[`planner`](docs\u002Ftools\u002Fplanner.md)** - 将复杂项目分解为结构化、可操作的计划\n- **[`consensus`](docs\u002Ftools\u002Fconsensus.md)** - 通过立场引导，从多个 AI 模型中获取专家意见\n\n**代码分析与质量**\n- **[`debug`](docs\u002Ftools\u002Fdebug.md)** - 系统性调查和根本原因分析\n- **[`precommit`](docs\u002Ftools\u002Fprecommit.md)** - 在提交前验证更改，防止回归\n- **[`codereview`](docs\u002Ftools\u002Fcodereview.md)** - 专业评审，带有严重程度等级和可操作反馈\n- **[`analyze`](docs\u002Ftools\u002Fanalyze.md)** *(默认禁用 - [启用](#tool-configuration))* - 理解整个代码库的架构、模式和依赖关系\n\n**开发工具** *(默认禁用 - [启用](#tool-configuration))*\n- **[`refactor`](docs\u002Ftools\u002Frefactor.md)** - 带有分解重点的智能代码重构\n- **[`testgen`](docs\u002Ftools\u002Ftestgen.md)** - 包含边缘情况的全面测试生成\n- **[`secaudit`](docs\u002Ftools\u002Fsecaudit.md)** - 基于 OWASP Top 10 的安全审计\n- **[`docgen`](docs\u002Ftools\u002Fdocgen.md)** - 生成带有复杂度分析的文档\n\n**实用工具**\n- **[`apilookup`](docs\u002Ftools\u002Fapilookup.md)** - 在子进程中强制查找当年的 API\u002FSDK 文档（节省当前上下文窗口内的 token），防止使用过时训练数据的回答\n- **[`challenge`](docs\u002Ftools\u002Fchallenge.md)** - 通过批判性分析，避免出现“你完全正确！”这样的回应\n- **[`tracer`](docs\u002Ftools\u002Ftracer.md)** *(默认禁用 - [启用](#tool-configuration))* - 用于调用流映射的静态分析提示\n\n\u003Cdetails>\n\u003Csummary>\u003Cb id=\"tool-configuration\">👉 工具配置\u003C\u002Fb>\u003C\u002Fsummary>\n\n### 默认配置\n\n为优化上下文窗口的使用，仅启用必要的工具：\n\n**默认启用：**\n- `chat`、`thinkdeep`、`planner`、`consensus` - 核心协作工具\n- `codereview`、`precommit`、`debug` - 必要的代码质量工具\n- `apilookup` - 快速查找 API\u002FSDK 信息\n- `challenge` - 批判性思维工具\n\n**默认禁用：**\n- `analyze`、`refactor`、`testgen`、`secaudit`、`docgen`、`tracer`\n\n### 启用更多工具\n\n要启用更多工具，请将其从 `DISABLED_TOOLS` 列表中移除：\n\n**选项 1：编辑 .env 文件**\n```bash\n# 默认配置（来自 .env.example）\nDISABLED_TOOLS=analyze,refactor,testgen,secaudit,docgen,tracer\n\n# 要启用特定工具，将其从列表中移除\n# 示例：启用 analyze 工具\nDISABLED_TOOLS=refactor,testgen,secaudit,docgen,tracer\n\n# 要启用所有工具\nDISABLED_TOOLS=\n```\n\n**选项 2：在 MCP 设置中配置**\n```json\n\u002F\u002F 在 ~\u002F.claude\u002Fsettings.json 或 .mcp.json 中\n{\n  \"mcpServers\": {\n    \"pal\": {\n      \"env\": {\n        \u002F\u002F 工具配置\n        \"DISABLED_TOOLS\": \"refactor,testgen,secaudit,docgen,tracer\",\n        \"DEFAULT_MODEL\": \"pro\",\n        \"DEFAULT_THINKING_MODE_THINKDEEP\": \"high\",\n        \n        \u002F\u002F API 配置\n        \"GEMINI_API_KEY\": \"your-gemini-key\",\n        \"OPENAI_API_KEY\": \"your-openai-key\",\n        \"OPENROUTER_API_KEY\": \"your-openrouter-key\",\n        \n        \u002F\u002F 日志记录和性能\n        \"LOG_LEVEL\": \"INFO\",\n        \"CONVERSATION_TIMEOUT_HOURS\": \"6\",\n        \"MAX_CONVERSATION_TURNS\": \"50\"\n      }\n    }\n  }\n}\n```\n\n**选项 3：启用所有工具**\n```json\n\u002F\u002F 将 DISABLED_TOOLS 删除或清空以启用所有功能\n{\n  \"mcpServers\": {\n    \"pal\": {\n      \"env\": {\n        \"DISABLED_TOOLS\": \"\"\n      }\n    }\n  }\n}\n```\n\n**注意：**\n- 必要的工具（`version`、`listmodels`）无法被禁用\n- 更改工具配置后，需重启 Claude 会话才能使更改生效\n- 每个工具都会增加上下文窗口的使用量，因此请仅启用您需要的工具\n\n\u003C\u002Fdetails>\n\n## 📺 观看工具实际运行效果\n\n\u003Cdetails>\n\u003Csummary>\u003Cb>聊天工具\u003C\u002Fb> - 协作式决策与多轮对话\u003C\u002Fsummary>\n\n**选择 Redis 还是 Memcached：**\n\n[Chat Redis or Memcached_web.webm](https:\u002F\u002Fgithub.com\u002Fuser-attachments\u002Fassets\u002F41076cfe-dd49-4dfc-82f5-d7461b34705d)\n\n**带有延续的多轮对话：**\n\n[Chat With Gemini_web.webm](https:\u002F\u002Fgithub.com\u002Fuser-attachments\u002Fassets\u002F37bd57ca-e8a6-42f7-b5fb-11de271e95db)\n\n\u003C\u002Fdetails>\n\n\u003Cdetails>\n\u003Csummary>\u003Cb>共识工具\u003C\u002Fb> - 多模型辩论与决策\u003C\u002Fsummary>\n\n**多模型共识辩论：**\n\n[PAL Consensus Debate](https:\u002F\u002Fgithub.com\u002Fuser-attachments\u002Fassets\u002F76a23dd5-887a-4382-9cf0-642f5cf6219e)\n\n\u003C\u002Fdetails>\n\n\u003Cdetails>\n\u003Csummary>\u003Cb>PreCommit 工具\u003C\u002Fb> - 全面的变更验证\u003C\u002Fsummary>\n\n**Pre-commit 验证工作流：**\n\n\u003Cdiv align=\"center\">\n  \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002FBeehiveInnovations_pal-mcp-server_readme_045584c73c01.png\" width=\"950\">\n\u003C\u002Fdiv>\n\n\u003C\u002Fdetails>\n\n\u003Cdetails>\n\u003Csummary>\u003Cb>API 查找工具\u003C\u002Fb> - 当前与过时的 API 文档\u003C\u002Fsummary>\n\n**没有 PAL - 过时的 APIs：**\n\n[API without PAL](https:\u002F\u002Fgithub.com\u002Fuser-attachments\u002Fassets\u002F01a79dc9-ad16-4264-9ce1-76a56c3580ee)\n\n**有 PAL - 当前的 APIs：**\n\n[API with PAL](https:\u002F\u002Fgithub.com\u002Fuser-attachments\u002Fassets\u002F5c847326-4b66-41f7-8f30-f380453dce22)\n\n\u003C\u002Fdetails>\n\n\u003Cdetails>\n\u003Csummary>\u003Cb>挑战工具\u003C\u002Fb> - 批判性思维 vs 本能性认同\u003C\u002Fsummary>\n\n**没有 PAL：**\n\n![without_pal@2x](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002FBeehiveInnovations_pal-mcp-server_readme_c19c869c8ff6.png)\n\n**有 PAL：**\n\n![with_pal@2x](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002FBeehiveInnovations_pal-mcp-server_readme_855b6613d596.png)\n\n\u003C\u002Fdetails>\n\n## 核心功能\n\n**AI 编排**\n- **自动模型选择** - Claude 会为每项任务挑选合适的 AI 模型\n- **多模型工作流** - 在单次对话中串联不同模型\n- **对话连续性** - 上下文在不同工具和模型间得以保留\n- **[上下文恢复](docs\u002Fcontext-revival.md)** - 即使上下文重置，也能继续对话\n\n**模型支持**\n- **多家提供商** - Gemini、OpenAI、Azure、X.AI、OpenRouter、DIAL、Ollama\n- **最新模型** - GPT-5、Gemini 3.0 Pro、O3、Grok-4、本地 Llama\n- **[思考模式](docs\u002Fadvanced-usage.md#thinking-modes)** - 控制推理深度与成本之间的平衡\n- **视觉支持** - 分析图片、图表、截图\n\n**开发者体验**\n- **引导式工作流** - 系统化的调查可避免仓促分析\n- **智能文件处理** - 自动展开目录，管理令牌限制\n- **网络搜索集成** - 访问最新文档和最佳实践\n- **[大提示支持](docs\u002Fadvanced-usage.md#working-with-large-prompts)** - 超越 MCP 的 25K 令牌限制\n\n## 示例工作流\n\n**多模型代码审查：**\n```\n“使用 Gemini Pro 和 O3 进行代码审查，然后用规划器制定修复策略”\n```\n→ Claude 系统地审查代码 → 咨询 Gemini Pro → 获取 O3 的观点 → 制定统一的行动计划\n\n**协作式调试：**\n```\n“以最大思考模式调试这个竞态条件，然后用 Precommit 验证修复结果”\n```\n→ 深入调查 → 专家分析 → 实施解决方案 → Pre-commit 验证\n\n**架构规划：**\n```\n“规划我们的微服务迁移，让 Pro 和 O3 就方案达成共识”\n```\n→ 结构化规划 → 多方专家意见 → 达成共识 → 制定实施路线图\n\n👉 **[高级使用指南](docs\u002Fadvanced-usage.md)** 提供复杂工作流、模型配置及高级用户功能的指导\n\n## 快速链接\n\n**📖 文档**\n- [文档概览](docs\u002Findex.md) - 主要指南的高层级地图\n- [入门指南](docs\u002Fgetting-started.md) - 完整的设置指南\n- [工具参考](docs\u002Ftools\u002F) - 包含示例的所有工具\n- [高级使用](docs\u002Fadvanced-usage.md) - 高级用户功能\n- [配置](docs\u002Fconfiguration.md) - 环境变量、限制条件\n- [添加提供商](docs\u002Fadding_providers.md) - 各提供商特定的设置（OpenAI、Azure、自定义网关）\n- [模型排名指南](docs\u002Fmodel_ranking.md) - 智能评分如何驱动自动模式建议\n\n**🔧 设置与支持**\n- [WSL 设置](docs\u002Fwsl-setup.md) - 适用于 Windows 用户\n- [故障排除](docs\u002Ftroubleshooting.md) - 常见问题\n- [贡献指南](docs\u002Fcontributions.md) - 代码标准、PR 流程\n\n## 许可证\n\nApache 2.0 许可证 - 详情请参阅 [LICENSE](LICENSE) 文件。\n\n## 致谢\n\n基于 **多模型 AI** 协作的强大能力构建 🤝\n- 来自真实人类的 **A**ctual **I**ntelligence\n- [MCP（模型上下文协议）](https:\u002F\u002Fmodelcontextprotocol.com)\n- [Codex CLI](https:\u002F\u002Fdevelopers.openai.com\u002Fcodex\u002Fcli)\n- [Claude Code](https:\u002F\u002Fclaude.ai\u002Fcode)\n- [Gemini](https:\u002F\u002Fai.google.dev\u002F)\n- [OpenAI](https:\u002F\u002Fopenai.com\u002F)\n- [Azure OpenAI](https:\u002F\u002Flearn.microsoft.com\u002Fazure\u002Fai-services\u002Fopenai\u002F)\n\n### 星标历史\n\n[![星标历史图表](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002FBeehiveInnovations_pal-mcp-server_readme_a41d3b421544.png)](https:\u002F\u002Fwww.star-history.com\u002F#BeehiveInnovations\u002Fpal-mcp-server&Date)","# PAL MCP Server 快速上手指南\n\nPAL MCP (Provider Abstraction Layer) 是一个模型上下文协议服务器，旨在将您喜爱的 AI CLI 工具（如 Claude Code、Codex CLI、Cursor 等）与多个 AI 模型连接起来。它支持在单个对话线程中协调多个模型进行协作、代码审查和复杂任务处理。\n\n## 环境准备\n\n在开始之前，请确保您的系统满足以下要求：\n\n*   **操作系统**: Linux, macOS 或 Windows (WSL2)\n*   **Python**: 版本 3.10 或更高\n*   **Git**: 已安装并配置\n*   **uv 包管理器**: 必须安装 [uv](https:\u002F\u002Fdocs.astral.sh\u002Fuv\u002Fgetting-started\u002Finstallation\u002F) (推荐用于快速安装和环境管理)\n*   **API 密钥**: 至少准备一个以下服务的 API Key：\n    *   [OpenRouter](https:\u002F\u002Fopenrouter.ai\u002F) (推荐：通过一个接口访问多种模型)\n    *   [Google Gemini](https:\u002F\u002Fmakersuite.google.com\u002Fapp\u002Fapikey)\n    *   [OpenAI](https:\u002F\u002Fplatform.openai.com\u002Fapi-keys)\n    *   [Ollama](https:\u002F\u002Follama.ai\u002F) (本地运行，免费)\n    *   其他支持的服务：Azure OpenAI, X.AI (Grok), DIAL 等\n\n## 安装步骤\n\n您可以选择以下两种方式之一进行安装：\n\n### 方式 A：克隆仓库并自动配置（推荐）\n\n此方式会自动处理依赖安装、配置文件生成以及常见 CLI 工具（Claude Desktop, Claude Code, Gemini CLI 等）的集成配置。\n\n```bash\n# 1. 克隆仓库\ngit clone https:\u002F\u002Fgithub.com\u002FBeehiveInnovations\u002Fpal-mcp-server.git\ncd pal-mcp-server\n\n# 2. 运行安装脚本\n# 该脚本会自动设置环境、读取系统环境变量中的 API Key，\n# 并自动配置 Claude Desktop, Claude Code, Gemini CLI, Codex CLI 等\n.\u002Frun-server.sh\n```\n*提示：运行前请确保已将 API Key 设置为系统环境变量，或在生成的 `.env` 文件中配置。*\n\n### 方式 B：使用 uvx 即时安装（无需克隆）\n\n如果您希望直接通过 `mcp.json` 或 `settings.json` 配置，无需下载源码，可使用此方式。\n\n1.  打开您的 MCP 配置文件 (例如 `~\u002F.claude\u002Fsettings.json` 或项目根目录下的 `.mcp.json`)。\n2.  添加以下配置内容（请替换 `your-key-here` 为您的实际 API Key）：\n\n```json\n{\n  \"mcpServers\": {\n    \"pal\": {\n      \"command\": \"bash\",\n      \"args\": [\"-c\", \"for p in $(which uvx 2>\u002Fdev\u002Fnull) $HOME\u002F.local\u002Fbin\u002Fuvx \u002Fopt\u002Fhomebrew\u002Fbin\u002Fuvx \u002Fusr\u002Flocal\u002Fbin\u002Fuvx uvx; do [ -x \\\"$p\\\" ] && exec \\\"$p\\\" --from git+https:\u002F\u002Fgithub.com\u002FBeehiveInnovations\u002Fpal-mcp-server.git pal-mcp-server; done; echo 'uvx not found' >&2; exit 1\"],\n      \"env\": {\n        \"PATH\": \"\u002Fusr\u002Flocal\u002Fbin:\u002Fusr\u002Fbin:\u002Fbin:\u002Fopt\u002Fhomebrew\u002Fbin:~\u002F.local\u002Fbin\",\n        \"GEMINI_API_KEY\": \"your-key-here\",\n        \"DISABLED_TOOLS\": \"analyze,refactor,testgen,secaudit,docgen,tracer\",\n        \"DEFAULT_MODEL\": \"auto\"\n      }\n    }\n  }\n}\n```\n\n## 基本使用\n\n安装完成后，您可以在支持的 AI CLI 工具（如 Claude Code, Cursor, Codex CLI）中直接通过自然语言调用 PAL 的功能。\n\n### 核心功能示例\n\n**1. 多模型协作分析**\n让不同的模型分别执行分析和优化任务：\n```text\nUse pal to analyze this code for security issues with gemini pro\n```\n```text\nDebug this error with o3 and then get flash to suggest optimizations\n```\n\n**2. 制定共识策略**\n让多个模型共同讨论并制定计划：\n```text\nPlan the migration strategy with pal, get consensus from multiple models\n```\n\n**3. 使用 Clink 桥接外部 CLI (子代理)**\n在当前会话中启动隔离的子代理执行特定角色任务（如规划、代码审查），保持主上下文干净：\n```text\nclink with cli_name=\"gemini\" role=\"planner\" to draft a phased rollout plan\n```\n```text\nclink with codex codereviewer to audit auth module for security issues\n```\n\n### 推荐工作流\n*   **主控制器**: 使用 Sonnet 4.5 或 GPT-5.2 Codex 作为主要编排者。\n*   **深度思考\u002F审查**: 调用 Gemini 3.0 Pro 或 O3 进行深度代码审查、调试和预提交验证。\n*   **上下文延续**: 当主模型上下文重置时，可利用其他模型的记忆“唤醒”主模型的上下文，实现无缝继续。\n\n> **注意**: 部分高级工具默认可能处于禁用状态以节省上下文窗口。如需启用所有功能，请修改配置中的 `DISABLED_TOOLS` 参数或参考完整文档进行自定义。","某全栈开发者正在为一个关键金融模块进行代码重构与安全审计，需要在有限时间内兼顾代码质量、安全性及多模型视角的验证。\n\n### 没有 pal-mcp-server 时\n- **模型切换繁琐**：为了对比不同模型的见解，开发者必须在 Claude Code、Gemini CLI 和 Codex 之间手动切换终端会话，导致工作流频繁中断。\n- **上下文严重割裂**：每次切换模型都需重新粘贴代码背景和之前的讨论记录，不仅耗时，还容易因信息遗漏导致模型产生幻觉或误判。\n- **难以形成共识**：无法让不同模型在同一语境下“辩论”技术方案（如选择何种加密算法），只能依靠人工整理各方观点，效率极低。\n- **主会话污染**：执行耗时的深度代码审查会占用大量上下文窗口，导致主开发会话的记忆容量被无关细节填满，影响后续任务。\n\n### 使用 pal-mcp-server 后\n- **统一编排多模型**：开发者只需在一个 Claude Code 会话中，即可通过 `clink` 命令直接调用 Gemini 或 Codex 作为子代理，实现“一键调度”整个 AI 团队。\n- **上下文无缝流转**：主会话的完整背景自动传递给子代理，例如让 Codex 专门审计安全漏洞后，直接将报告返回给主会话，无需任何手动复制粘贴。\n- **自动化协作辩论**：利用 conversation threading 功能，让 GPT-5 和 Gemini 针对架构方案进行多轮对话辩论，最终输出综合最优解供开发者决策。\n- **隔离式并行任务**：启动独立的 `codereviewer` 角色子代理在隔离环境中运行，既完成了深度扫描，又完美保持了主开发窗口的清爽与专注。\n\npal-mcp-server 将分散的单一模型能力整合为协同工作的\"AI 开发团队”，在保持开发者控制权的同时，极大提升了复杂任务的解决深度与效率。","https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002FBeehiveInnovations_pal-mcp-server_c19c869c.png","BeehiveInnovations","Beehive Innovations","https:\u002F\u002Foss.gittoolsai.com\u002Favatars\u002FBeehiveInnovations_88c62f74.png","",null,"BusyMac","https:\u002F\u002Fwww.busymac.com","https:\u002F\u002Fgithub.com\u002FBeehiveInnovations",[81,85,89,93],{"name":82,"color":83,"percentage":84},"Python","#3572A5",93.7,{"name":86,"color":87,"percentage":88},"Shell","#89e051",3.2,{"name":90,"color":91,"percentage":92},"PowerShell","#012456",3.1,{"name":94,"color":95,"percentage":96},"Dockerfile","#384d54",0.1,11414,977,"2026-04-12T15:47:14","NOASSERTION","Linux, macOS, Windows","非必需（仅在本地运行 Ollama 等模型时需要，取决于具体模型需求）","未说明",{"notes":105,"python":106,"dependencies":107},"该工具是一个 MCP 服务器，主要用于连接不同的 AI CLI 工具和模型提供商。核心依赖是包管理器 'uv' 和 'git'。若使用云端模型（如 OpenAI, Gemini, Claude 等），无需本地 GPU；若选择通过 Ollama 运行本地模型，则需根据所选模型配置相应的本地硬件资源。支持通过 .env 文件配置多种 API Key。","3.10+",[108,109],"uv","git",[13,52],"2026-03-27T02:49:30.150509","2026-04-13T07:03:59.618347",[114,119,124,129,134,139],{"id":115,"question_zh":116,"answer_zh":117,"source_url":118},31474,"为什么 Zen MCP 经常会卡住或无限运行？","这通常是由特定的 AI 模型（特别是 o3 模型）引起的。用户反馈表明，切换到 o3-mini 模型可以解决此问题。此外，尝试关闭“灵活处理”（flex processing）选项也可能有助于提高稳定性，因为该选项可能会导致处理速度变慢并引发卡顿。","https:\u002F\u002Fgithub.com\u002FBeehiveInnovations\u002Fpal-mcp-server\u002Fissues\u002F181",{"id":120,"question_zh":121,"answer_zh":122,"source_url":123},31475,"使用 OpenRouter 模型时遇到“请求超时”错误怎么办？","目前无法通过设置环境变量（如 TIMEOUT 或 MCP_SERVER_REQUEST_TIMEOUT）来解决此问题，因为超时是由调用方（如某些 IDE 的 stdio 连接）控制的，而非服务器端。例如，Claude Code 可以等待更长时间，但其他集成可能在 1 分钟后强制断开。如果使用的 IDE 不支持长耗时响应，可能需要考虑更换支持更长超时时间的开发环境。","https:\u002F\u002Fgithub.com\u002FBeehiveInnovations\u002Fpal-mcp-server\u002Fissues\u002F233",{"id":125,"question_zh":126,"answer_zh":127,"source_url":128},31476,"使用 Gemini 模型时出现\"Response blocked or incomplete\"错误的原因是什么？","这通常是因为使用了免费的 Gemini API Key。免费密钥不保证服务的稳定性，其可用性受负载、时间段等因素影响极大，失败率较高。如果频繁遇到此错误，建议检查是否正在使用免费密钥，并考虑升级到付费层级以获得更稳定的服务。","https:\u002F\u002Fgithub.com\u002FBeehiveInnovations\u002Fpal-mcp-server\u002Fissues\u002F251",{"id":130,"question_zh":131,"answer_zh":132,"source_url":133},31477,"如何通过 uvx 运行 zen-mcp-server 时正确加载 .env 文件？","当通过 uvx 运行时，服务器可能无法自动从安装目录加载 .env 文件。解决方法是在 MCP 配置文件中显式定义所需的环境变量。不要依赖自动加载，而是直接在配置对象的 \"env\" 字段中填入 API Key，例如：\"env\": { \"OPENROUTER_API_KEY\": \"sk-or-...\" }。","https:\u002F\u002Fgithub.com\u002FBeehiveInnovations\u002Fpal-mcp-server\u002Fissues\u002F185",{"id":135,"question_zh":136,"answer_zh":137,"source_url":138},31478,"是否可以在一次调用中咨询多个 AI 模型以获取综合观点？","是的，该功能已实现。现在支持在一次调用中同时使用多个模型（如 gemini-pro, o3, grok-3）。系统会并行发送请求，并返回包含各模型独立响应及可选的“聚合洞察”（aggregated insights）的结果。这有助于在规划或调试时快速获取多样化的 unbiased 观点。","https:\u002F\u002Fgithub.com\u002FBeehiveInnovations\u002Fpal-mcp-server\u002Fissues\u002F53",{"id":140,"question_zh":141,"answer_zh":142,"source_url":128},31479,"如何确认当前使用的 API Key 类型是否导致了不稳定？","如果遇到频繁的随机错误或响应被阻断，首先检查使用的是否为免费 API Key（特别是 Gemini）。维护者指出，免费密钥的表现就像“迷你彩票”，随时可能因负载过高而失败。如果错误发生在请求刚开始且频率很高（如 50%），极大概率是免费密钥的限制所致，而非代码逻辑问题。",[144,149,154,159,164,169,174,179,184,189,194,199,204,209,214,219,224,229,234,239],{"id":145,"version":146,"summary_zh":147,"released_at":148},231204,"v9.8.2","## v9.8.2 (2025-12-15)\n\n### 错误修复\n\n- 允许通过 `is_dangerous_path()` 检查主目录的子目录 ([`e5548ac`](https:\u002F\u002Fgithub.com\u002FBeehiveInnovations\u002Fpal-mcp-server\u002Fcommit\u002Fe5548acb984ca4f8b2ae8381f879a0285094257f))\n\n- 路径遍历漏洞 - 在 `is_dangerous_path()` 中使用前缀匹配 ([`9ed15f4`](https:\u002F\u002Fgithub.com\u002FBeehiveInnovations\u002Fpal-mcp-server\u002Fcommit\u002F9ed15f405a9462b4db7aa44ca2d989e092c008e4))\n\n- 使用 `Path.is_relative_to()` 进行跨平台危险路径检测 ([`91ffb51`](https:\u002F\u002Fgithub.com\u002FBeehiveInnovations\u002Fpal-mcp-server\u002Fcommit\u002F91ffb51564e5655ec91111938039ed81e0d8e4c6))\n\n- **安全**: 处理 macOS 上的符号链接系统目录 ([`ba08308`](https:\u002F\u002Fgithub.com\u002FBeehiveInnovations\u002Fpal-mcp-server\u002Fcommit\u002Fba08308a23d1c1491099c5d0eae548077bd88f9f))\n\n### 杂项\n\n- 将版本号同步到 `config.py` [跳过 CI] ([`c492735`](https:\u002F\u002Fgithub.com\u002FBeehiveInnovations\u002Fpal-mcp-server\u002Fcommit\u002Fc4927358720277efa0373b339bd8e06ee06498d0))\n\n---\n\n**详细变更**: [v9.8.1...v9.8.2](https:\u002F\u002Fgithub.com\u002FBeehiveInnovations\u002Fpal-mcp-server\u002Fcompare\u002Fv9.8.1...v9.8.2)","2025-12-15T17:07:27",{"id":150,"version":151,"summary_zh":152,"released_at":153},231205,"v9.8.1","## v9.8.1（2025-12-15）\n\n### Bug 修复\n\n- **providers**: 为 OpenRouter 响应端点省略 store 参数 ([`1f8b58d`](https:\u002F\u002Fgithub.com\u002FBeehiveInnovations\u002Fpal-mcp-server\u002Fcommit\u002F1f8b58d607c2809b9fa78860718a69207cb66e32))\n\n### 杂项\n\n- 将版本号同步到 config.py [跳过 CI] ([`69a42a7`](https:\u002F\u002Fgithub.com\u002FBeehiveInnovations\u002Fpal-mcp-server\u002Fcommit\u002F69a42a71d19d66f1d94d51fa27db29323e3d9a63))\n\n### 重构\n\n- **tests**: 处理代码审查反馈 ([`0c3e63c`](https:\u002F\u002Fgithub.com\u002FBeehiveInnovations\u002Fpal-mcp-server\u002Fcommit\u002F0c3e63c0c7f1556f4b6686f9c6f30e4bb4a48c7c))\n\n- **tests**: 移除未使用的 setUp 方法 ([`b6a8d68`](https:\u002F\u002Fgithub.com\u002FBeehiveInnovations\u002Fpal-mcp-server\u002Fcommit\u002Fb6a8d682d920c2283724b588818bc1162a865d74))\n\n---\n\n**详细变更**: [v9.8.0...v9.8.1](https:\u002F\u002Fgithub.com\u002FBeehiveInnovations\u002Fpal-mcp-server\u002Fcompare\u002Fv9.8.0...v9.8.1)","2025-12-15T11:53:54",{"id":155,"version":156,"summary_zh":157,"released_at":158},231206,"v9.8.0","## v9.8.0 (2025-12-15)\n\n### 杂项\n\n- 将版本号同步到 config.py [跳过 CI] ([`cb97a89`](https:\u002F\u002Fgithub.com\u002FBeehiveInnovations\u002Fpal-mcp-server\u002Fcommit\u002Fcb97a891dec6ab7c56b8b35c277ab3680af384d9))\n\n### 功能\n\n- 通过 OpenRouter 添加 Claude Opus 4.5 模型 ([`813ce5c`](https:\u002F\u002Fgithub.com\u002FBeehiveInnovations\u002Fpal-mcp-server\u002Fcommit\u002F813ce5c9f7db2910eb12d8c84d3d99f464c430ed))\n\n### 测试\n\n- 为 Opus 4.5 别名添加全面的测试覆盖 ([`cf63fd2`](https:\u002F\u002Fgithub.com\u002FBeehiveInnovations\u002Fpal-mcp-server\u002Fcommit\u002Fcf63fd25440d599f2ec006bb8cfda5b8a6f61524))\n\n---\n\n**详细变更**: [v9.7.0...v9.8.0](https:\u002F\u002Fgithub.com\u002FBeehiveInnovations\u002Fpal-mcp-server\u002Fcompare\u002Fv9.7.0...v9.8.0)","2025-12-15T11:38:23",{"id":160,"version":161,"summary_zh":162,"released_at":163},231207,"v9.7.0","## v9.7.0 (2025-12-15)\n\n### 杂项\n\n- 将版本同步到 config.py [跳过 CI] ([`aa85644`](https:\u002F\u002Fgithub.com\u002FBeehiveInnovations\u002Fpal-mcp-server\u002Fcommit\u002Faa85644c9b15893443107c3a62ec58cd7b9dc532))\n\n### 功能\n\n- 使用正确的 --enable 标志重新启用 Clink Codex 的网页搜索 ([`e7b9f3a`](https:\u002F\u002Fgithub.com\u002FBeehiveInnovations\u002Fpal-mcp-server\u002Fcommit\u002Fe7b9f3a5d7e06c690c82b9fd13a93310bcf388ed))\n\n---\n\n**详细变更**: [v9.6.0...v9.7.0](https:\u002F\u002Fgithub.com\u002FBeehiveInnovations\u002Fpal-mcp-server\u002Fcompare\u002Fv9.6.0...v9.7.0)","2025-12-15T10:22:27",{"id":165,"version":166,"summary_zh":167,"released_at":168},231208,"v9.6.0","## v9.6.0 (2025-12-15)\n\n### 杂项\n\n- 将版本号同步到 config.py [跳过 CI] ([`94ff26c`](https:\u002F\u002Fgithub.com\u002FBeehiveInnovations\u002Fpal-mcp-server\u002Fcommit\u002F94ff26c673a64087eb29f8f54c1828f1157c594a))\n\n### 功能\n\n- 支持原生安装的 Claude CLI 检测 ([`adc6231`](https:\u002F\u002Fgithub.com\u002FBeehiveInnovations\u002Fpal-mcp-server\u002Fcommit\u002Fadc6231b98886f0bc35cb04d04d948eba2f0f058))\n\n---\n\n**详细变更**: [v9.5.0...v9.6.0](https:\u002F\u002Fgithub.com\u002FBeehiveInnovations\u002Fpal-mcp-server\u002Fcompare\u002Fv9.5.0...v9.6.0)","2025-12-15T10:18:55",{"id":170,"version":171,"summary_zh":172,"released_at":173},231209,"v9.5.0","## v9.5.0 (2025-12-11)\n\n### Bug 修复\n\n- Grok 测试 ([`39c7721`](https:\u002F\u002Fgithub.com\u002FBeehiveInnovations\u002Fpal-mcp-server\u002Fcommit\u002F39c77215e5d6892269e523ff25b706dd5671c042))\n\n### 杂项任务\n\n- 将版本号同步到 config.py [跳过 CI] ([`5c3dd75`](https:\u002F\u002Fgithub.com\u002FBeehiveInnovations\u002Fpal-mcp-server\u002Fcommit\u002F5c3dd75ca6b259f590bfd5078ea8e2f684e52de4))\n\n- 将版本号同步到 config.py [跳过 CI] ([`605633b`](https:\u002F\u002Fgithub.com\u002FBeehiveInnovations\u002Fpal-mcp-server\u002Fcommit\u002F605633b2a2b044bbc5e41f2994dde27409a5b9b4))\n\n### 文档\n\n- 清理 ([`74f26e8`](https:\u002F\u002Fgithub.com\u002FBeehiveInnovations\u002Fpal-mcp-server\u002Fcommit\u002F74f26e82e7a9c8a0214deef1cb18a3b2fa074050))\n\n- 清理 ([`2b22174`](https:\u002F\u002Fgithub.com\u002FBeehiveInnovations\u002Fpal-mcp-server\u002Fcommit\u002F2b221746fee6f7749d8aed8d07a85e428ac8e00f))\n\n- 更新子标题 ([`591287c`](https:\u002F\u002Fgithub.com\u002FBeehiveInnovations\u002Fpal-mcp-server\u002Fcommit\u002F591287cb2f442a1fa34cd1139e3a0ad887388e5b))\n\n### 新特性\n\n- 支持 GPT-5.2 ([`8b16405`](https:\u002F\u002Fgithub.com\u002FBeehiveInnovations\u002Fpal-mcp-server\u002Fcommit\u002F8b16405f0609e232ff808361dc2a4d8ec258b0f3))\n\n- 支持 Grok-4.1 https:\u002F\u002Fgithub.com\u002FBeehiveInnovations\u002Fpal-mcp-server\u002Fissues\u002F339 ([`514c9c5`](https:\u002F\u002Fgithub.com\u002FBeehiveInnovations\u002Fpal-mcp-server\u002Fcommit\u002F514c9c58fcc91933348d2188ed8c82bbe98132f2))\n\n---\n\n**详细变更**: [v9.4.2...v9.5.0](https:\u002F\u002Fgithub.com\u002FBeehiveInnovations\u002Fpal-mcp-server\u002Fcompare\u002Fv9.4.2...v9.5.0)","2025-12-11T20:15:00",{"id":175,"version":176,"summary_zh":177,"released_at":178},231210,"v9.4.2","## v9.4.2 (2025-12-04)\n\n### 错误修复\n\n- 品牌重塑，详情请参阅 [docs\u002Fname-change.md](docs\u002Fname-change.md)（[`b2dc849`](https:\u002F\u002Fgithub.com\u002FBeehiveInnovations\u002Fpal-mcp-server\u002Fcommit\u002Fb2dc84992d70839b29b611178b3871f4922b747f)）\n\n### 杂项\n\n- 将版本号同步到 config.py [跳过 CI]（[`bcfacce`](https:\u002F\u002Fgithub.com\u002FBeehiveInnovations\u002Fpal-mcp-server\u002Fcommit\u002Fbcfaccecd490859fe189f45df4cf5b8e102d7874)）\n\n---\n\n**详细变更**: [v9.4.1...v9.4.2](https:\u002F\u002Fgithub.com\u002FBeehiveInnovations\u002Fpal-mcp-server\u002Fcompare\u002Fv9.4.1...v9.4.2)","2025-12-04T14:16:03",{"id":180,"version":181,"summary_zh":182,"released_at":183},231211,"v9.4.1","## v9.4.1 (2025-11-21)\n\n### 错误修复\n\n- 修复回归问题 https:\u002F\u002Fgithub.com\u002FBeehiveInnovations\u002Fzen-mcp-server\u002Fissues\u002F338 ([`aceddb6`](https:\u002F\u002Fgithub.com\u002FBeehiveInnovations\u002Fzen-mcp-server\u002Fcommit\u002Faceddb655fc36918108b3da1f926bdd4e94875a2))\n\n### 杂项\n\n- 将版本号同步到 config.py [跳过 CI] ([`c4461a4`](https:\u002F\u002Fgithub.com\u002FBeehiveInnovations\u002Fzen-mcp-server\u002Fcommit\u002Fc4461a466fab9c647b0a5035328c4d0f3e28f647))\n\n---\n\n**详细变更**: [v9.4.0...v9.4.1](https:\u002F\u002Fgithub.com\u002FBeehiveInnovations\u002Fzen-mcp-server\u002Fcompare\u002Fv9.4.0...v9.4.1)","2025-11-21T05:32:22",{"id":185,"version":186,"summary_zh":187,"released_at":188},231212,"v9.4.0","## v9.4.0 (2025-11-18)\n\n### 错误修复\n\n- 修复 gemini 3.0 pro open router 的测试失败问题 ([`19a2a89`](https:\u002F\u002Fgithub.com\u002FBeehiveInnovations\u002Fzen-mcp-server\u002Fcommit\u002F19a2a89b12c5dec53aea21a4244aff7796a5e049))\n\n### 杂项\n\n- 将版本号同步到 config.py [跳过 CI] ([`d3de61f`](https:\u002F\u002Fgithub.com\u002FBeehiveInnovations\u002Fzen-mcp-server\u002Fcommit\u002Fd3de61f8787ab60261d09f2c7f362c50d2093799))\n\n### 功能\n\n- 为 Open Router 提供 Gemini 3.0 Pro 预览版 ([`bbfdfac`](https:\u002F\u002Fgithub.com\u002FBeehiveInnovations\u002Fzen-mcp-server\u002Fcommit\u002Fbbfdfac511668e8ae60f9b9b5d41eb9ab55d74cf))\n\n### 重构\n\n- 在 codex CLI 中启用搜索功能 ([`1579d9f`](https:\u002F\u002Fgithub.com\u002FBeehiveInnovations\u002Fzen-mcp-server\u002Fcommit\u002F1579d9f806a653bb04c9c73ab304cdd0e78fbdfa))\n\n---\n\n**详细变更**: [v9.3.1...v9.4.0](https:\u002F\u002Fgithub.com\u002FBeehiveInnovations\u002Fzen-mcp-server\u002Fcompare\u002Fv9.3.1...v9.4.0)","2025-11-18T16:51:34",{"id":190,"version":191,"summary_zh":192,"released_at":193},231213,"v9.3.1","## v9.3.1 (2025-11-18)\n\n### 杂项\n\n- 将版本号同步到 config.py [跳过 CI] ([`d256098`](https:\u002F\u002Fgithub.com\u002FBeehiveInnovations\u002Fzen-mcp-server\u002Fcommit\u002Fd2560983402abf084608f7750f05407a8d3e20a0))\n\n---\n\n**详细变更**: [v9.3.0...v9.3.1](https:\u002F\u002Fgithub.com\u002FBeehiveInnovations\u002Fzen-mcp-server\u002Fcompare\u002Fv9.3.0...v9.3.1)\n","2025-11-18T16:33:00",{"id":195,"version":196,"summary_zh":197,"released_at":198},231214,"v9.3.0","## v9.3.0 (2025-11-18)\n\n### Chores\n\n- Sync version to config.py [skip ci] ([`3748d47`](https:\u002F\u002Fgithub.com\u002FBeehiveInnovations\u002Fzen-mcp-server\u002Fcommit\u002F3748d47faba7d871f2dd379f2c8646aa8cd3c6e9))\n\n---\n\n**Detailed Changes**: [v9.2.2...v9.3.0](https:\u002F\u002Fgithub.com\u002FBeehiveInnovations\u002Fzen-mcp-server\u002Fcompare\u002Fv9.2.2...v9.3.0)\n","2025-11-18T16:29:20",{"id":200,"version":201,"summary_zh":202,"released_at":203},231215,"v9.2.2","## v9.2.2 (2025-11-18)\n\n### Bug Fixes\n\n- **build**: Include clink resources in package ([`e9ac1ce`](https:\u002F\u002Fgithub.com\u002FBeehiveInnovations\u002Fzen-mcp-server\u002Fcommit\u002Fe9ac1ce3354fbb124a72190702618f94266b8459))\n\n### Chores\n\n- Sync version to config.py [skip ci] ([`749bc73`](https:\u002F\u002Fgithub.com\u002FBeehiveInnovations\u002Fzen-mcp-server\u002Fcommit\u002F749bc7307949fa0b0e026bfcfbd546d7619eba8b))\n\n---\n\n**Detailed Changes**: [v9.2.1...v9.2.2](https:\u002F\u002Fgithub.com\u002FBeehiveInnovations\u002Fzen-mcp-server\u002Fcompare\u002Fv9.2.1...v9.2.2)\n","2025-11-18T07:34:07",{"id":205,"version":206,"summary_zh":207,"released_at":208},231216,"v9.2.1","## v9.2.1 (2025-11-18)\n\n### Bug Fixes\n\n- **server**: Iterate provider instances during shutdown ([`d40fc83`](https:\u002F\u002Fgithub.com\u002FBeehiveInnovations\u002Fzen-mcp-server\u002Fcommit\u002Fd40fc83d7549293372f3d20cc599a79ec355acef))\n\n### Chores\n\n- Sync version to config.py [skip ci] ([`84f6c4f`](https:\u002F\u002Fgithub.com\u002FBeehiveInnovations\u002Fzen-mcp-server\u002Fcommit\u002F84f6c4fb241257b611f4b954c22a6b9340007a73))\n\n---\n\n**Detailed Changes**: [v9.2.0...v9.2.1](https:\u002F\u002Fgithub.com\u002FBeehiveInnovations\u002Fzen-mcp-server\u002Fcompare\u002Fv9.2.0...v9.2.1)\n","2025-11-18T07:06:26",{"id":210,"version":211,"summary_zh":212,"released_at":213},231217,"v9.2.0","## v9.2.0 (2025-11-18)\n\n### Chores\n\n- Sync version to config.py [skip ci] ([`7a1de64`](https:\u002F\u002Fgithub.com\u002FBeehiveInnovations\u002Fzen-mcp-server\u002Fcommit\u002F7a1de6477aae88bfe7a2f677faf0794169651354))\n\n### Documentation\n\n- Streamline advanced usage guide by reorganizing table of contents for improved navigation ([`698d391`](https:\u002F\u002Fgithub.com\u002FBeehiveInnovations\u002Fzen-mcp-server\u002Fcommit\u002F698d391b26a0dd565eada8bfa6e67e549ce1dd20))\n\n- Update .env.example to include new GPT-5.1 model options and clarify existing model descriptions ([`dbbfef2`](https:\u002F\u002Fgithub.com\u002FBeehiveInnovations\u002Fzen-mcp-server\u002Fcommit\u002Fdbbfef292c67ed54f90f7612c9c14d4095bd6c45))\n\n- Update advanced usage and configuration to include new GPT-5.1 models and enhance tool parameters ([`807c9df`](https:\u002F\u002Fgithub.com\u002FBeehiveInnovations\u002Fzen-mcp-server\u002Fcommit\u002F807c9df70e3b54031ec6beea10f3975455b36dfb))\n\n### Features\n\n- Add new GPT-5.1 models to configuration files and update model selection logic in OpenAI provider ([`8e9aa23`](https:\u002F\u002Fgithub.com\u002FBeehiveInnovations\u002Fzen-mcp-server\u002Fcommit\u002F8e9aa2304d5e9ea9a9f8dc2a13a27a1ced6b1608))\n\n- Enhance model support by adding GPT-5.1 to .gitignore and updating cassette maintenance documentation for dual-model testing ([`f713d8a`](https:\u002F\u002Fgithub.com\u002FBeehiveInnovations\u002Fzen-mcp-server\u002Fcommit\u002Ff713d8a354a37c32a806c98994e6f949ecd64237))\n\n---\n\n**Detailed Changes**: [v9.1.4...v9.2.0](https:\u002F\u002Fgithub.com\u002FBeehiveInnovations\u002Fzen-mcp-server\u002Fcompare\u002Fv9.1.4...v9.2.0)\n","2025-11-18T06:50:18",{"id":215,"version":216,"summary_zh":217,"released_at":218},231218,"v9.1.4","## v9.1.4 (2025-11-18)\n\n### Bug Fixes\n\n- Replaced deprecated Codex web search configuration ([`2ec64ba`](https:\u002F\u002Fgithub.com\u002FBeehiveInnovations\u002Fzen-mcp-server\u002Fcommit\u002F2ec64ba7489acc586846b25eedf94a4f05d5bd2d))\n\n### Chores\n\n- Sync version to config.py [skip ci] ([`4d3d177`](https:\u002F\u002Fgithub.com\u002FBeehiveInnovations\u002Fzen-mcp-server\u002Fcommit\u002F4d3d177d91370097ca7ac4f922fa3a8b69ce3250))\n\n---\n\n**Detailed Changes**: [v9.1.3...v9.1.4](https:\u002F\u002Fgithub.com\u002FBeehiveInnovations\u002Fzen-mcp-server\u002Fcompare\u002Fv9.1.3...v9.1.4)\n","2025-11-18T06:38:20",{"id":220,"version":221,"summary_zh":222,"released_at":223},231219,"v9.1.3","## v9.1.3 (2025-10-22)\n\n### Bug Fixes\n\n- Reduced token usage, removed parameters from schema that CLIs never seem to use ([`3e27319`](https:\u002F\u002Fgithub.com\u002FBeehiveInnovations\u002Fzen-mcp-server\u002Fcommit\u002F3e27319e60b0287df918856b58b2bbf042c948a8))\n\n- Telemetry option no longer available in gemini 0.11 ([`2a8dff0`](https:\u002F\u002Fgithub.com\u002FBeehiveInnovations\u002Fzen-mcp-server\u002Fcommit\u002F2a8dff0cc8a3f33111533cdb971d654637ed0578))\n\n### Chores\n\n- Sync version to config.py [skip ci] ([`9e163f9`](https:\u002F\u002Fgithub.com\u002FBeehiveInnovations\u002Fzen-mcp-server\u002Fcommit\u002F9e163f9dc0654fc28961c9897b7c787a2b96e57d))\n\n- Sync version to config.py [skip ci] ([`557e443`](https:\u002F\u002Fgithub.com\u002FBeehiveInnovations\u002Fzen-mcp-server\u002Fcommit\u002F557e443a63ffd733fb41faaa8696f6f4bb2c2fd1))\n\n### Refactoring\n\n- Improved precommit system prompt ([`3efff60`](https:\u002F\u002Fgithub.com\u002FBeehiveInnovations\u002Fzen-mcp-server\u002Fcommit\u002F3efff6056e322ee1531d7bed5601038c129a8b29))\n\n---\n\n**Detailed Changes**: [v9.1.2...v9.1.3](https:\u002F\u002Fgithub.com\u002FBeehiveInnovations\u002Fzen-mcp-server\u002Fcompare\u002Fv9.1.2...v9.1.3)\n","2025-10-22T16:35:19",{"id":225,"version":226,"summary_zh":227,"released_at":228},231220,"v9.1.2","## v9.1.2 (2025-10-21)\n\n### Bug Fixes\n\n- Configure codex with a longer timeout ([`d2773f4`](https:\u002F\u002Fgithub.com\u002FBeehiveInnovations\u002Fzen-mcp-server\u002Fcommit\u002Fd2773f488af28986632846652874de9ff633049c))\n\n- Handle claude's array style JSON https:\u002F\u002Fgithub.com\u002FBeehiveInnovations\u002Fzen-mcp-server\u002Fissues\u002F295 ([`d5790a9`](https:\u002F\u002Fgithub.com\u002FBeehiveInnovations\u002Fzen-mcp-server\u002Fcommit\u002Fd5790a9bfef719f03d17f2d719f1882e55d13b3b))\n\n### Chores\n\n- Sync version to config.py [skip ci] ([`04132f1`](https:\u002F\u002Fgithub.com\u002FBeehiveInnovations\u002Fzen-mcp-server\u002Fcommit\u002F04132f1459f1e086afd8e3d456f671b63338f846))\n\n---\n\n**Detailed Changes**: [v9.1.1...v9.1.2](https:\u002F\u002Fgithub.com\u002FBeehiveInnovations\u002Fzen-mcp-server\u002Fcompare\u002Fv9.1.1...v9.1.2)\n","2025-10-21T06:41:58",{"id":230,"version":231,"summary_zh":232,"released_at":233},231221,"v9.1.1","## v9.1.1 (2025-10-17)\n\n### Bug Fixes\n\n- Failing test ([`aed3e3e`](https:\u002F\u002Fgithub.com\u002FBeehiveInnovations\u002Fzen-mcp-server\u002Fcommit\u002Faed3e3ee80c440ac8ab0d4abbf235b84df723d18))\n\n- Handler for parsing multiple generated code blocks ([`f4c20d2`](https:\u002F\u002Fgithub.com\u002FBeehiveInnovations\u002Fzen-mcp-server\u002Fcommit\u002Ff4c20d2a20e1c57d8b10e8f508e07e2a8d72f94a))\n\n- Improved error reporting; codex cli would at times fail to figure out how to handle plain-text \u002F JSON errors ([`95e69a7`](https:\u002F\u002Fgithub.com\u002FBeehiveInnovations\u002Fzen-mcp-server\u002Fcommit\u002F95e69a7cb234305dcd37dcdd2f22be715922e9a8))\n\n### Chores\n\n- Sync version to config.py [skip ci] ([`942757a`](https:\u002F\u002Fgithub.com\u002FBeehiveInnovations\u002Fzen-mcp-server\u002Fcommit\u002F942757a360a74c021b2a1aa63e394f18f5abcecd))\n\n---\n\n**Detailed Changes**: [v9.1.0...v9.1.1](https:\u002F\u002Fgithub.com\u002FBeehiveInnovations\u002Fzen-mcp-server\u002Fcompare\u002Fv9.1.0...v9.1.1)\n","2025-10-17T20:33:54",{"id":235,"version":236,"summary_zh":237,"released_at":238},231222,"v9.1.0","## v9.1.0 (2025-10-17)\n\n### Chores\n\n- Sync version to config.py [skip ci] ([`3ee0c8f`](https:\u002F\u002Fgithub.com\u002FBeehiveInnovations\u002Fzen-mcp-server\u002Fcommit\u002F3ee0c8f555cb51b975700290919c2a8e2ada8cc4))\n\n### Features\n\n- Enhance review prompts to emphasize static analysis ([`36e66e2`](https:\u002F\u002Fgithub.com\u002FBeehiveInnovations\u002Fzen-mcp-server\u002Fcommit\u002F36e66e2e9a44a73a466545d4d3477ecb2cb3e669))\n\n---\n\n**Detailed Changes**: [v9.0.4...v9.1.0](https:\u002F\u002Fgithub.com\u002FBeehiveInnovations\u002Fzen-mcp-server\u002Fcompare\u002Fv9.0.4...v9.1.0)\n","2025-10-17T05:23:43",{"id":240,"version":241,"summary_zh":242,"released_at":243},231223,"v9.0.4","## v9.0.4 (2025-10-17)\n\n### Chores\n\n- Sync version to config.py [skip ci] ([`8c6f653`](https:\u002F\u002Fgithub.com\u002FBeehiveInnovations\u002Fzen-mcp-server\u002Fcommit\u002F8c6f6532d843f7f1b283ce9b6472e5ba991efe16))\n\n---\n\n**Detailed Changes**: [v9.0.3...v9.0.4](https:\u002F\u002Fgithub.com\u002FBeehiveInnovations\u002Fzen-mcp-server\u002Fcompare\u002Fv9.0.3...v9.0.4)\n","2025-10-17T04:45:58"]