# Zikra MCP server

> Provides a shared, governed memory layer for AI agents and teams using PostgreSQL and pgvector for persistent, multi-project context.

- Repository: https://github.com/getzikra/Zikra
- Page: https://hostedmcp.org/servers/getzikra/zikra
- Transport: —
- Stars: 9
- License: MIT

## Environment variables

| Name | Required | Secret | Description |
|---|---|---|---|
| `DB_BACKEND` | no | no |  |
| `DB_HOST` | no | no |  |
| `DB_NAME` | no | no |  |
| `DB_PASSWORD` | no | no |  |
| `DB_PORT` | no | no |  |
| `DB_USER` | no | no |  |

## FAQ

### How does Zikra handle memory for teams?

It provides a shared memory pool with role-based access control (RBAC) including owner, admin, developer, and viewer roles per project.

### Can I host Zikra myself?

Yes, Zikra is designed to be self-hosted with no cloud dependency, using PostgreSQL or SQLite as a backend.

### Does it support multiple AI tools?

Yes, it works across multiple AI tools such as Claude Code, Gemini CLI, and Codex.

### How does the memory hygiene work?

It uses a built-in hygiene prompt that manages confidence decay, orphan detection, and stale cleanup.
