# agent-memory MCP server

> Provides persistent, typed, and cross-runtime memory for AI agents using a local SQLite database.

- Repository: https://github.com/UnityInFlow/agent-memory
- Page: https://hostedmcp.org/servers/unityinflow/agent-memory
- Transport: —
- Stars: 0
- License: MIT

## Environment variables

| Name | Required | Secret | Description |
|---|---|---|---|
| `AGENT_MEMORY_DB` | no | no |  |
| `AGENT_MEMORY_OLLAMA_URL` | no | no |  |

## Tools (17)

- `memory_store`
- `memory_search`
- `memory_list`
- `memory_update`
- `memory_link`
- `memory_unlink`
- `memory_forget`
- `search_mode`
- `content`
- `tags`
- `scope`
- `ttl_secs`
- `kind`
- `relates_to`
- `supersedes`
- `caused_by`
- `limit`

## FAQ

### Is my data stored in the cloud?

No, the server is local and offline. Your memory never leaves your machine because it uses an embedded SQLite database.

### Can different AI agents share the same memory?

Yes, any MCP-speaking runtime pointed at the same database file shares the same memory, allowing decisions recorded by one agent to be recalled by another.

### How does the search function work if I don't have Ollama?

If Ollama is unavailable, the server automatically falls back to SQLite FTS5 keyword ranking.

### Are memories permanently deleted when they decay?

No, memories only fade in ranking over time due to exponential decay; they are only removed via explicit TTL expiry or a manual forget command.
