# MCP Memory Service — Persistent Agent Memory API

> Persistent memory service for agent pipelines: REST API + knowledge graph + consolidation, with MCP integration for Claude and friends.

## Install

Merge the JSON below into your `.mcp.json`:

## Quick Use

1. Install:
   ```bash
   pip install mcp-memory-service
   ```
2. Start an HTTP server (for local dev):
   ```bash
   MCP_ALLOW_ANONYMOUS_ACCESS=true memory server --http
   # API at http://localhost:8000
   ```
3. Store a memory (example):
   ```bash
   curl -sS -X POST http://localhost:8000/api/memories \
     -H 'Content-Type: application/json' \
     -H 'X-Agent-ID: researcher' \
     -d '{"content":"API rate limit is 100 req/min","tags":["api","limits"]}'
   ```

## Intro

Persistent memory service for agent pipelines: REST API + knowledge graph + consolidation, with MCP integration for Claude and friends.

- **Best for:** teams building agent workflows that need shared, queryable, long-lived memory beyond one chat session
- **Works with:** any HTTP client; common agent frameworks; MCP-capable clients (optional)
- **Setup time:** 20–45 minutes

## Practical Notes

- Quant: the project calls out a framework-agnostic REST API with 76 endpoints—start by standardizing 5–10 “memory ops” your agents must use.
- Quant: track retrieval precision by running the same query across 10 tasks and measuring how often the top-3 hits are actually used.

---

## How to design memory you can trust

Treat memory as *data you can audit*, not magical recall:

- Store short, atomic facts with tags (source, system, owner, environment).
- Always include an `X-Agent-ID` strategy so retrieval can be scoped.
- Separate “write” and “read” phases: write after the task, read before the task.

## A simple workflow pattern

1. **Task start:** query with a narrow tag set.
2. **During work:** write only verified decisions (not raw speculation).
3. **Task end:** consolidate duplicates and add a short summary memory.

## Safety defaults

If you enable anonymous access for local testing, keep it strictly on localhost and disable it in any shared environment.

### FAQ

**Q: Is this only for MCP clients?**
A: No. You can treat it as a plain HTTP memory service; MCP integration is optional.

**Q: How do I avoid noisy memories?**
A: Write only verified facts and decisions, and require tags + agent scope for every write.

**Q: What should I measure first?**
A: Retrieval precision and reuse rate: how often retrieved items are actually used in the next task.

## Source & Thanks

> Source: https://github.com/doobidoo/mcp-memory-service
> License: Apache-2.0
> GitHub stars: 1,835 · forks: 278

---

<!-- ZH -->

## 快速使用

1. 安装：
   ```bash
   pip install mcp-memory-service
   ```
2. 启动 HTTP 服务（本地开发示例）：
   ```bash
   MCP_ALLOW_ANONYMOUS_ACCESS=true memory server --http
   # API: http://localhost:8000
   ```
3. 写入一条记忆（示例）：
   ```bash
   curl -sS -X POST http://localhost:8000/api/memories \
     -H 'Content-Type: application/json' \
     -H 'X-Agent-ID: researcher' \
     -d '{"content":"API rate limit is 100 req/min","tags":["api","limits"]}'
   ```

## 简介

面向 Agent 流水线的持久化记忆服务：提供 REST API、知识图谱与自动归并能力，可接入 MCP/Claude 等工具链，适配 LangGraph/CrewAI 等框架，把上下文沉淀为可检索、可复用的长期资产。

- **适合谁：** 需要跨任务共享、可检索、可长期沉淀记忆的 Agent 团队与工作流
- **可搭配：** 任何 HTTP 客户端；常见 agent 框架；可选 MCP 客户端
- **准备时间：** 20–45 分钟

## 实战建议

- 量化建议：项目说明 REST API 覆盖 76 个 endpoints；先选出 5–10 个必须统一的“记忆操作”再扩展。
- 量化建议：把同一查询在 10 个任务里复跑，统计 top-3 命中内容被实际采用的比例（检索精度）。

## 如何把记忆做成“可相信的数据”

把记忆当作**可审计的数据层**，而不是“玄学召回”：

- 存储短小、原子化事实，并用 tags 标注来源/系统/所有者/环境。
- 设计好 `X-Agent-ID` 规范，确保检索可以按 agent 身份隔离。
- 把“写入”和“读取”拆成两个阶段：任务前读，任务后写。

## 一个可落地的流程模板

1. **任务开始：** 用少量关键 tags 做窄检索。
2. **执行过程中：** 只写入已验证的决策与结论（不要把猜测当事实）。
3. **任务结束：** 做一次去重/归并，再补一条可复用的总结记忆。

## 安全默认

匿名访问只用于本地测试；任何共享环境都应关闭，并加上鉴权与网络隔离。

### FAQ

**只能给 MCP 客户端用吗？**
答：不是。它也可以作为普通 HTTP 记忆服务使用；MCP 集成是可选项。

**怎么避免记忆变成噪音？**
答：只写入已验证事实/决策；每次写入都要求 tags + agent 作用域。

**优先量化什么？**
答：检索精度与复用率：被检索出来的内容在下一次任务中被真正采用的比例。

## 来源与感谢

> Source: https://github.com/doobidoo/mcp-memory-service
> License: Apache-2.0
> GitHub stars: 1,835 · forks: 278


---
Source: https://tokrepo.com/en/workflows/mcp-memory-service-persistent-agent-memory-api
Author: MCP Hub