Agent-VC MCP Server

Agent-VC MCP Server

Provides AI agents with persistent state, version control, and task management capabilities powered by Fossil SCM. It enables agents to manage files in a sandboxed environment with full commit history and built-in ticket tracking.

Category
访问服务器

README

Agent-VC MCP Server

CI

A specialized MCP server for AI agents to have persistent state, version control, and task management using Fossil SCM.

Why Agent-VC?

  • Persistent State: Agents can store files in a structured, versioned workspace.
  • Reproducibility: Full history of every commit, allowing rollbacks to any state.
  • Task Management: Built-in tickets/tasks (via Fossil's ticket system) to track progress without external dependencies.
  • Sandboxed: Operations are strictly limited to ~/.agent-vc/workspace/.

Structure & File Management

Core Directories

  • ~/.agent-vc/: The root directory for all Agent-VC data and configuration.
  • ~/.agent-vc/version_control.fossil: The Fossil repository file. This is a single SQLite database that stores your entire project history, version control objects, and task/ticket data.
  • ~/.agent-vc/workspace/: The active checkout directory. All operations performed by the agent occur within this directory. This is the only place files are read from or written to.

File and Folder Mapping

Agent-VC uses a "Project" based organization within the shared workspace. This allows the agent to manage multiple logical projects or modules within a single version-controlled repository.

Concept Description Path Example
Workspace Root The base directory for all operations. ~/.agent-vc/workspace/
Project Folder A subdirectory representing a specific project. ~/.agent-vc/workspace/chatbot-v1/
Project Subpath A file or nested folder within a project. ~/.agent-vc/workspace/chatbot-v1/src/utils.js

Working with the MCP Tools

When using the agent-vc tools, you interact with files using the project and subPath parameters. This abstraction maps directly to the filesystem:

  1. Initializing a Project: Running vc_init_project(name: "my-app") creates the directory ~/.agent-vc/workspace/my-app/.
  2. Writing Files: Running write_file(project: "my-app", subPath: "main.py", content: "...") writes to ~/.agent-vc/workspace/my-app/main.py.
  3. Nested Folders: You can use forward slashes in the subPath to create nested structures: write_file(project: "my-app", subPath: "docs/specs/readme.md", ...) will automatically create the docs/specs/ directories if they don't exist.

[!NOTE] All files created using the write_file tool are automatically added to Fossil's tracking (fossil add). To record these changes permanently in history, you must call vc_commit.

Installation

  1. Install Fossil: Follow the official installation instructions on the Fossil website.

  2. Run via npx: You can run the server directly using npx:

    npx agent-vc-mcp
    
  3. Add to your MCP Config (e.g., claude_desktop_config.json or equivalent agent config):

{
  "mcpServers": {
    "agent-vc": {
      "command": "npx",
      "args": ["agent-vc-mcp"],
      "env": {
        "AGENT_VC_WORKSPACE": "/path/to/my/project",
        "AGENT_VC_FOSSIL_DB": "/path/to/backup.fossil"
      }
    }
  }
}

Environment Variables

You can tune Agent-VC behavior by setting these environment variables:

Variable Description Default
AGENT_VC_ROOT_OVERRIDE The root folder for agent data. ~/.agent-vc/
AGENT_VC_WORKSPACE Path to the workspace checkout. ~/.agent-vc/workspace/ (if cwd is / or .)
AGENT_VC_FOSSIL_DB Path to the Fossil repository file. ~/.agent-vc/version_control.fossil

[!TIP] Setting AGENT_VC_WORKSPACE is useful if you want the agent to operate directly on an existing local directory that you've already initialized with Fossil SCM.

Tools

Version Control

  • vc_init_project(name): Create a project subfolder.
  • vc_commit(message): Commit all changes in the workspace.
  • vc_status(): Get current status of tracking and changes.
  • vc_diff(): View changes in the workspace.

File Operations (Project Scoped)

  • write_file(project, subPath, content): Writes a file and automatically adds it to Fossil tracking.
  • read_file(project, subPath): Reads a file.
  • list_files(project): Lists all files in a project.

Task Management

  • task_create(title, description): Create a project task.
  • task_list(status?): List existing tasks.
  • task_update(id, status, comment?): Update task status.

Usage Examples

🏢 Initializing a Project

Create a new project folder structure in the workspace.

{
  "name": "vc_init_project",
  "arguments": { "name": "chatbot-v1" }
}

💾 Writing and Committing Files

Write code and commit it to history.

// 1. Write the file
{
  "name": "write_file",
  "arguments": {
    "project": "chatbot-v1",
    "subPath": "index.js",
    "content": "console.log('Hello Agent!');"
  }
}

// 2. Commit the changes
{
  "name": "vc_commit",
  "arguments": { "message": "initial commit for chatbot" }
}

📋 Managing Tasks (Fossil Tickets)

Track work using the built-in ticket system.

// 1. Create a task
{
  "name": "task_create",
  "arguments": {
    "title": "Add streaming support",
    "description": "Implement chunked response parsing."
  }
}

// 2. List open tasks (to get the Ticket ID)
{
  "name": "task_list",
  "arguments": { "status": "Open" }
}

// 3. Update a task (using the first 7-10 characters of the Ticket ID)
{
  "name": "task_update",
  "arguments": {
    "id": "8d0aa38",
    "status": "Closed",
    "comment": "Work completed and tested."
  }
}

🔍 Inspecting Progress & History

// Check modified files
{ "name": "vc_status", "arguments": {} }

// View code diff
{ "name": "vc_diff", "arguments": {} }

// List all files in a project
{
  "name": "list_files",
  "arguments": { "project": "chatbot-v1" }
}

// Read a specific file
{
  "name": "read_file",
  "arguments": {
    "project": "chatbot-v1",
    "subPath": "index.js"
  }
}

Further Documentation

For more advanced Fossil commands and features (like branching, merging, and syncing), refer to the official Fossil SCM Documentation.

推荐服务器

Baidu Map

Baidu Map

百度地图核心API现已全面兼容MCP协议,是国内首家兼容MCP协议的地图服务商。

官方
精选
JavaScript
Playwright MCP Server

Playwright MCP Server

一个模型上下文协议服务器,它使大型语言模型能够通过结构化的可访问性快照与网页进行交互,而无需视觉模型或屏幕截图。

官方
精选
TypeScript
Magic Component Platform (MCP)

Magic Component Platform (MCP)

一个由人工智能驱动的工具,可以从自然语言描述生成现代化的用户界面组件,并与流行的集成开发环境(IDE)集成,从而简化用户界面开发流程。

官方
精选
本地
TypeScript
Audiense Insights MCP Server

Audiense Insights MCP Server

通过模型上下文协议启用与 Audiense Insights 账户的交互,从而促进营销洞察和受众数据的提取和分析,包括人口统计信息、行为和影响者互动。

官方
精选
本地
TypeScript
VeyraX

VeyraX

一个单一的 MCP 工具,连接你所有喜爱的工具:Gmail、日历以及其他 40 多个工具。

官方
精选
本地
graphlit-mcp-server

graphlit-mcp-server

模型上下文协议 (MCP) 服务器实现了 MCP 客户端与 Graphlit 服务之间的集成。 除了网络爬取之外,还可以将任何内容(从 Slack 到 Gmail 再到播客订阅源)导入到 Graphlit 项目中,然后从 MCP 客户端检索相关内容。

官方
精选
TypeScript
Kagi MCP Server

Kagi MCP Server

一个 MCP 服务器,集成了 Kagi 搜索功能和 Claude AI,使 Claude 能够在回答需要最新信息的问题时执行实时网络搜索。

官方
精选
Python
e2b-mcp-server

e2b-mcp-server

使用 MCP 通过 e2b 运行代码。

官方
精选
Neon MCP Server

Neon MCP Server

用于与 Neon 管理 API 和数据库交互的 MCP 服务器

官方
精选
Exa MCP Server

Exa MCP Server

模型上下文协议(MCP)服务器允许像 Claude 这样的 AI 助手使用 Exa AI 搜索 API 进行网络搜索。这种设置允许 AI 模型以安全和受控的方式获取实时的网络信息。

官方
精选