# Stash — Self-Hosted Memory + MCP SSE on Postgres/pgvector

> Stash is a self-hosted memory layer for agents with an MCP SSE endpoint, backed by Postgres + pgvector and a consolidation pipeline for long-lived recall.

## Install

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

## Quick Use

1. Start the stack:
   ```bash
   git clone https://github.com/alash3al/stash.git
   cd stash
   cp .env.example .env
   docker compose up
   ```
2. Point your MCP client to the SSE endpoint:
   - URL: `http://localhost:8080/sse`
3. Verify: ask the agent to write one memory, then query it back by tags or keywords.

## Intro

Stash is a self-hosted memory layer for agents with an MCP SSE endpoint, backed by Postgres + pgvector and a consolidation pipeline for long-lived recall.

- **Best for:** teams that want persistent memory for agents without relying on a hosted SaaS, plus an MCP-compatible interface
- **Works with:** Docker Compose, PostgreSQL + pgvector, MCP clients that support SSE (Cursor, Claude Desktop, Windsurf)
- **Setup time:** 15-35 minutes

## Practical Notes

- Quant: the README highlights an 8-stage consolidation pipeline—treat that as a controllable knob for memory quality vs cost.
- Quant: use one SSE endpoint per environment (dev/staging/prod) so memory doesn’t cross-contaminate across contexts.

## Why it matters

Stash is interesting when your agent needs long-term memory that stays inside your infra, and you want an MCP endpoint that multiple clients can reuse.

- Quick start is one `docker compose up`, which makes it easy to pilot in a team.
- SSE transport means multiple MCP-capable clients can point at the same service endpoint.
- Positions memory as a consolidation pipeline (episodes → facts), which maps to how teams debug memory quality.

## Rollout pattern

- Start with one agent and one project; define what counts as “write-worthy” memory (facts, decisions, constraints).
- Add retention and consolidation schedules once you observe retrieval precision in real tasks.
- Create a “memory schema” (tags, owners, environments) so retrieval stays scoped and predictable.

## Watchouts

Memory is also a data-leak vector. Treat the store as sensitive: enforce auth, avoid storing secrets, and keep environments separated.

### FAQ

**Q: Is this only for Cursor?**
A: No. It works with any MCP client that supports SSE endpoints; the README lists multiple clients.

**Q: How do I keep memory useful?**
A: Write only verified facts/decisions and enforce tags + scope per agent or per project.

**Q: Can I run it in production?**
A: Yes, but add auth, backups, and monitoring; don’t leave it anonymously accessible on a network.

## Source & Thanks

> Source: https://github.com/alash3al/stash
> License: Apache-2.0
> GitHub stars: 687 · forks: 30

---

<!-- ZH -->

## 快速使用

1. 启动服务：
   ```bash
   git clone https://github.com/alash3al/stash.git
   cd stash
   cp .env.example .env
   docker compose up
   ```
2. 在 MCP 客户端里配置 SSE：
   - URL：`http://localhost:8080/sse`
3. 验证：让 Agent 写入一条记忆，再按关键词/标签检索回来。

## 简介

Stash 是自托管的 Agent 记忆层，基于 Postgres + pgvector 存储，并通过 SSE 暴露 MCP Server；搭配后台归并流程把对话“片段”沉淀为可检索、可复用的长期记忆资产。

- **适合谁：** 想要自托管持久化记忆层（非 SaaS），并希望通过 MCP 接入到各类客户端的团队
- **可搭配：** Docker Compose、PostgreSQL + pgvector、支持 MCP SSE 的客户端（Cursor / Claude Desktop / Windsurf）
- **准备时间：** 15-35 分钟

## 实战建议

- 量化信息：README 提到 8 阶段归并流水线；可把它当作“记忆质量 vs 成本”的可调旋钮。
- 量化信息：建议 dev/staging/prod 各自一套 SSE endpoint，避免记忆跨环境污染。

## 为什么值得收录

当你想把记忆层留在自家基础设施里，并希望多个 MCP 客户端共享同一个“记忆服务端点”时，Stash 这类实现会很有用。

- 一条 `docker compose up` 就能跑起来，便于团队快速试点。
- SSE 传输意味着多个 MCP 客户端可以共用同一服务端点。
- 把记忆当作“归并流水线”，有利于排查记忆质量问题。

## 落地路径

- 先限定 1 个 agent + 1 个项目，明确哪些信息值得写入（事实/决策/约束）。
- 观察真实任务中的检索命中率后，再配置保留期与归并调度。
- 建立“记忆 schema”（tags/owner/environment），确保检索可控且不串线。

## 注意事项

记忆层也可能成为数据泄露路径。建议强制鉴权、避免存秘密，并把环境严格隔离。

### FAQ

**只能给 Cursor 用吗？**
答：不是。任何支持 MCP SSE 的客户端都能用；README 也列了多种客户端。

**怎么让记忆更有用？**
答：只写已验证事实/决策，并要求 tags + 作用域（按 agent 或项目隔离）。

**能上生产吗？**
答：可以，但要加鉴权、备份与监控；不要把它匿名暴露在网络上。

## 来源与感谢

> Source: https://github.com/alash3al/stash
> License: Apache-2.0
> GitHub stars: 687 · forks: 30


---
Source: https://tokrepo.com/en/workflows/stash-self-hosted-memory-mcp-sse-on-postgres-pgvector
Author: MCP Hub