# Agent Integration Guides
mcp-memory-service provides persistent shared memory for multi-agent systems via a **framework-agnostic REST API**. No MCP client library required — any HTTP client works.
## Integration Method by Framework
| Framework | Recommended Method | Transport |
|---|---|---|
| LangGraph | REST API via `httpx` | HTTP |
| CrewAI | REST API via `httpx` | HTTP |
| AutoGen | REST API via `httpx` | HTTP |
| Any HTTP client | REST API | HTTP |
| Claude Desktop / Code | MCP protocol | stdio / SSE |
All guides below use the REST API (port 8000 by default). The MCP protocol is for Claude Desktop/Code only.
## Quick Start
```bash
pip install mcp-memory-service
MCP_ALLOW_ANONYMOUS_ACCESS=true memory server --http
# Server running at http://localhost:8000
```
## Tag Convention for Agent Scoping
Use the `agent:` namespace to scope memories by agent identity:
| Tag pattern | Purpose |
|---|---|
| `agent:<id>` | Memories owned by a specific agent (`agent:researcher`) |
| `crew:<name>` | Memories shared within a crew (`crew:analysis-team`) |
| `proj:<name>` | Project-scoped memories (`proj:myproject`) |
**Auto-tagging via HTTP header:** Send `X-Agent-ID: <id>` in any store request — the server automatically appends `agent:<id>` to the memory's tags.
```bash
curl -X POST http://localhost:8000/api/memories \
-H "Content-Type: application/json" \
-H "X-Agent-ID: researcher" \
-d '{"content": "Rate limit is 100 req/min", "tags": ["api"]}'
# Stored with tags: ["api", "agent:researcher"]
```
## Framework-Specific Guides
- [LangGraph](langgraph.md) — Memory nodes in StateGraph, cross-graph sharing
- [CrewAI](crewai.md) — Custom tools, agent-scoped and crew-scoped memory
- [AutoGen](autogen.md) — Context injection, function tool schema, conversation dedup
- [HTTP Generic](http-generic.md) — All 15 REST endpoints, auth patterns, async examples
## Key Differentiators vs Alternatives
| | Mem0 | Zep | DIY Redis+Pinecone | **mcp-memory-service** |
|---|---|---|---|---|
| License | Proprietary | Enterprise | — | **Apache 2.0** |
| Cost | Per-call API | Enterprise | Infra costs | **$0** |
| Knowledge graph | No | Limited | No | **Yes (typed edges)** |
| Auto consolidation | No | No | No | **Yes (decay + compression)** |
| On-premise embeddings | No | No | Manual | **Yes (ONNX, local)** |
| Privacy | Cloud | Cloud | Partial | **100% local** |
| Hybrid search | No | Yes | Manual | **Yes (BM25 + vector)** |
| MCP protocol | No | No | No | **Yes** |
| REST API | Yes | Yes | Manual | **Yes (15 endpoints)** |
