# Metacortex MCP server

> Provides a serverless persistent memory service for AI agents using Firestore vector search to save, search, and retrieve durable context.

- Repository: https://github.com/chrishonson/Metacortex
- Page: https://hostedmcp.org/servers/chrishonson/metacortex
- Transport: —
- Stars: 2
- License: MIT

## Environment variables

| Name | Required | Secret | Description |
|---|---|---|---|
| `MCP_ADMIN_TOKEN` | no | no |  |
| `MCP_BASE_URL` | no | no |  |
| `MCP_CLIENT_PROFILES_JSON` | no | no |  |
| `MCP_SMOKE_MODE` | no | no |  |

## Tools (3)

- `save_context`
- `search_context`
- `fetch_context`

## FAQ

### What can AI clients do with Metacortex?

Clients can search existing knowledge, save new durable memories from chat, and fetch full stored memory records using specific IDs.

### Does Metacortex store raw image files?

No, it does not store raw image bytes for download; instead, it normalizes images into retrieval text and allows you to link to external assets via artifact references.

### What are the requirements for deploying to Firebase?

Production deployment of Cloud Functions requires the Firebase Blaze plan.

### How does the search functionality work?

The search performs an exact metadata filter followed by a vector search using Firestore's findNearest function with cosine distance.