
Gitmem
io.github.gitmem-dev由 GitHub 账号 gitmem-dev 发布v1.2.1更新于 Oct 24, 2026
Persistent institutional memory for AI coding agents. Memory that compounds.
概览
GitMem 是一个本地 MCP 服务器,为 AI 编程助手提供跨会话的持久记忆,记录错误、成功经验、模式、决策和未完成工作。
- 可以做什么
- 在助手行动前检索以往会话中的相关经验,遇到类似任务时自动提示曾经的教训。把过程记录为教训(错误)、成功、模式、决策和未完成的线程,并为教训附上反方理由。在会话结束时进行结构化复盘,让上下文延续到下一次会话。搜索和管理已存记忆、跟踪未完成事项,并协调多个代理。通过命令行向导完成服务器与客户端配置,并合并进已有配置文件。
- 适用场景
- 适合在同一项目上长期使用 AI 编程助手、希望它不再重复已知错误并持续记住架构决策和未完成工作的场景。适用于 Claude Code、Cursor、VS Code、Windsurf 或其他兼容 MCP 的客户端。
- 运行要求
- 通过 npm 包 gitmem-mcp 以 stdio 在本地运行,需要 Node.js 和 npx。免费版把数据保存在本地 .gitmem/ 目录,无需账号。可选的 Pro 版自托管在自己的 Supabase 项目上,需要 SUPABASE_URL 和 SUPABASE_SERVICE_ROLE_KEY,以及用于激活的许可证密钥。
安装
在 SourceWeft 中
- 打开 控制台中的 Gitmem,将其添加到工作区。
- 为需要使用其工具的对话启用该服务。
Desktop only,通过 STDIO。 STDIO 服务会启动本地进程,因此需要 SourceWeft 桌面宿主。
其他 MCP 客户端
参照 仓库 中的启动说明。
README
[GitMem — Institutional memory for AI coding agents]
[npm version] [npm downloads] [MIT License] [Build] [Node.js >= 18]
Documentation · npm · Getting Started · Tool Reference
GitMem is an MCP server that gives your AI coding agent persistent learning memory across agent sessions. It remembers mistakes (scars), successes (wins), and decisions — so your agent learns from experience instead of starting from scratch every time.
What's MCP? Model Context Protocol is how AI coding tools connect to external capabilities. GitMem is an MCP server — install it once and your agent gains persistent memory.
Works with Claude Code, Cursor, VS Code (Copilot), Windsurf, and any MCP-compatible client.
Quick Start
One command. The wizard auto-detects your IDE and sets up everything:
.gitmem/directory with starter scars- MCP server config (
.mcp.json,.vscode/mcp.json,.cursor/mcp.json, etc.) - Instructions file (
CLAUDE.md,.cursorrules,.windsurfrules,.github/copilot-instructions.md) - Lifecycle hooks (where supported)
.gitignoreupdated
Already have existing config? The wizard merges without destroying anything. Re-running is safe.
How It Works
- Recall — Before acting, the agent checks memory for relevant lessons from past sessions
- Work — The agent does the task, applying past lessons automatically
- Learn — Mistakes become scars, successes become wins, strategies become patterns
- Close — Session reflection persists context for next time
Every scar includes counter-arguments — reasons why someone might reasonably ignore it. This prevents memory from becoming a pile of rigid rules.
What Gets Remembered
Key Features
- Automatic Recall — Scars surface before the agent takes similar actions
- Session Continuity — Context, threads, and rapport carry across sessions
- Closing Ceremony — Structured reflection captures what broke, what worked, and what to do differently
- 20+ MCP Tools — Full toolkit for memory management, search, threads, and multi-agent coordination
- Zero Config —
npx gitmem-mcp initand you're running - Non-Destructive — Merges with your existing
.mcp.json,CLAUDE.md, and hooks
Supported Clients
The wizard auto-detects your IDE. Use --client to override.
Manual MCP configuration
Add this to your MCP client's config file:
CLI Commands
Pro Tier
Self-hosted on your own Supabase. You bring the infrastructure, gitmem sets it up.
Quick start
The activate command creates all tables, views, RPC functions, and indexes automatically. No manual SQL needed.
See docs/pro-setup-guide.md for the full guide.
The free tier gives you everything for solo projects. Pro makes recall smarter and memory portable.
GitMem + MEMORY.md
Your AI agent likely has its own memory file (MEMORY.md, .cursorrules, etc.). Here's how they work together:
Tip: Include .gitmem/agent-briefing.md in your MEMORY.md for a lightweight bridge between the two systems.
Privacy & Data
- Local-first — All data stored in
.gitmem/on your machine by default - No telemetry — GitMem does not collect usage data or phone home
- Cloud opt-in — Pro tier Supabase backend requires explicit configuration via environment variables
- Your data — Sessions, scars, and decisions belong to you. Delete
.gitmem/to remove everything
Development
See CONTRIBUTING.md for full development setup.
License
MIT — see LICENSE.
来源:README.md,提交 2e3dee2
工具
0版本历史
1- v1.2.1最新Sep 16, 2026
评分
已满足适用检查项的 50%
该评分衡量此收录条目的完整程度,依据注册中心条目、端点与公开来源计算得出。它不是对服务器本身的评价,也不表示服务器安全或当前可用:可连通性是带日期的检查结果,评分也不包含安装量或用户评价。
- 近 30 天内可连通25 分
请保持端点能够响应 MCP 握手;SourceWeft 每 30 天探测一次。
- 已声明认证方式10 分
请在注册中心条目中声明凭据请求头,或保持端点能够响应探测,以便判定其认证方式。
- 已列出工具15 分
请让服务器在无凭据时响应 tools/list,或发布可在沙箱中无凭据启动的软件包。
- 描述必需15 分
- 图标10 分
- 文档15 分
- 许可证10 分