# Zekra

> Zekra is an open-source MCP server for knowledge and memory. Shared long-term memory and knowledge graph for AI agents, with citations and self-hosting. It runs as a hosted remote server, so there is nothing to install locally.

- **Kind:** MCP server
- **Pricing:** Open source
- **Platforms:** API, Web, self-hosted, iOS, Android, macOS, Windows, Linux, CLI
- **Website:** https://zekra.dev
- **Repository:** https://github.com/fadymondy/zekra
- **MCP Registry name:** io.github.fadymondy/zekra
- **Categories:** knowledge and memory
- **Listing updated:** 2026-11-15

## About

Most memory stores keep appending; Zekra decides on every write whether to add, update, invalidate or ignore, so a contradicted fact is replaced instead of duplicated. Forgetting soft-invalidates, so history stays queryable. Recall blends vector similarity with multilingual BM25, fuses them by reciprocal rank, reranks, and can expand one hop through the entity graph.

Memories are grouped into brains, one per project or team, with per-agent grants and tokens enforced on the server. Connectors for text, markdown, crawler, GitHub, SQL and webhook sources feed content through the same write pipeline. Other features named by the project include a chat agent that answers with citations, a secrets vault, knowledge-gap detection and an activity log.

The site also describes media memories: OCR for Arabic, English and other scripts, PDF text, video keyframes with transcripts, and audio transcription. It can build decks and PDF or Word reports from a brain's notes and translate them between Arabic and English. Zekra is AGPL-3.0, can be self-hosted with Docker Compose (about 6 GB of disk for models), or used from the hosted app. Media reading needs an optional OCR service when self-hosted.

## Features

- **Hybrid recall:** Combines vectors, BM25, reranking and one-hop graph expansion.
- **Write decisions:** Each write is added, updated, invalidated or ignored to avoid duplicates.
- **Brains with grants:** Namespaces per project or team with server-enforced tokens.
- **Knowledge graph:** Extracts entities and typed relations that can be traversed.
- **Data source connectors:** Pulls text, markdown, web, GitHub, SQL and webhook content.
- **Media memory and decks:** Reads images, PDFs, video and audio and builds Arabic-English decks.

## Good for

- Carry project decisions between Claude Code and Cursor sessions
- Index a GitHub repo or docs site as agent memory
- Turn scanned receipts and whiteboard photos into searchable text
- Draft a client deck from stored notes
- Run a private memory service for a team

## Install

### Claude Code (Terminal)

```sh
claude mcp add --transport http zekra https://mcp.zekra.dev
```

### Claude Desktop (Settings → Connectors → Add custom connector)

```sh
https://mcp.zekra.dev
```

### Cursor (~/.cursor/mcp.json)

```json
{
  "mcpServers": {
    "zekra": {
      "url": "https://mcp.zekra.dev"
    }
  }
}
```

### VS Code (.vscode/mcp.json)

```json
{
  "servers": {
    "zekra": {
      "type": "http",
      "url": "https://mcp.zekra.dev"
    }
  }
}
```

### Windsurf (~/.codeium/windsurf/mcp_config.json)

```json
{
  "mcpServers": {
    "zekra": {
      "serverUrl": "https://mcp.zekra.dev"
    }
  }
}
```

### Other clients (your client's MCP settings (most use this shape))

```json
{
  "mcpServers": {
    "zekra": {
      "type": "http",
      "url": "https://mcp.zekra.dev"
    }
  }
}
```

## Tools

- `memory_retain`: Store a memory.
- `memory_recall`: Recall memories for a query.
- `memory_get`: Fetch a memory.
- `memory_forget`: Soft-invalidate a memory.
- `memory_share`: Share a memory.
- `memory_edit`: Edit a memory.
- `presentation_create_from_outline`: Create a deck from an outline.
- `presentation_translate`: Translate a deck between Arabic and English.
- `presentation_export`: Export a deck to PDF.
- `presentation_share`: Create a read-only share link.
- `media_retain`: Remember an image, PDF, video or audio file.
- `media_get`: Return the text read from stored media.

## FAQ

### What licence is Zekra under?

AGPL-3.0, according to the site and README.

### Can I self-host it?

Yes. The README gives a Docker Compose setup using Postgres, a Go API and embedding and reranking servers.

### Which clients connect?

Claude Desktop and Code, Codex, Gemini and Cursor, through one CLI that installs the MCP connection.

### Does it delete memories?

Forgetting soft-invalidates them; history remains queryable.

### Is Zekra free?

Yes. Zekra is open source, so you can use and self-host it for free.

### How do I add Zekra to Claude Code?

Run this command in your terminal: claude mcp add --transport http zekra https://mcp.zekra.dev

### What are the best alternatives to Zekra?

The closest alternatives to Zekra are Codebase Memory, XERJ, Claude Code Ultimate Guide, Nautilus Compass Memory and note1.

## Alternatives

- [Codebase Memory](https://ai-powered.com/mcp/codebase-memory): Codebase Memory is an open-source MCP server for knowledge and memory.
- [XERJ](https://ai-powered.com/mcp/xerj): XERJ is an open-source MCP server for knowledge and memory.
- [Claude Code Ultimate Guide](https://ai-powered.com/mcp/claude-code-ultimate-guide): Claude Code Ultimate Guide is an open-source MCP server for ai models and llm tooling.
- [Nautilus Compass Memory](https://ai-powered.com/mcp/nautilus-compass-memory): Nautilus Compass Memory is an open-source MCP server for knowledge and memory.
- [note1](https://ai-powered.com/mcp/note1): note1 is an MCP server for productivity.
- [MedXpert 图书馆连接器（通用版）](https://ai-powered.com/mcp/medxpert): MedXpert 图书馆连接器（通用版） is an open-source MCP server for knowledge and memory.
- [Context Nest](https://ai-powered.com/mcp/context-nest): Context Nest is a freemium MCP server for knowledge and memory.
- [Tulsa Ecosystem Map](https://ai-powered.com/mcp/tulsa-ecosystem-map): Tulsa Ecosystem Map is a free MCP server for knowledge and memory.

---
Source: https://ai-powered.com/mcp/zekra
