# Omneum

> Omneum is an open-source MCP server for knowledge and memory. Context control plane that picks the most useful retrieved passages under a token budget. It runs locally on your machine and is started by your AI client.

- **Kind:** MCP server
- **Pricing:** Open source
- **Platforms:** macOS, Linux, CLI, self-hosted
- **Repository:** https://github.com/cole-h06/omneum
- **MCP Registry name:** io.github.cole-h06/omneum
- **Categories:** knowledge and memory, AI models and LLM tooling
- **Listing updated:** 2026-11-15

## About

Omneum targets context rot, where long agent contexts degrade answers even after the right information was retrieved. It estimates how reliable each source is and how well each claim is supported, tracking when several sources trace back to one upstream origin so repeated reports are not counted as independent proof.

The selection step keeps original passages, ranks them by claim coverage and, when a question is given, local lexical relevance, then fills the token budget. The README's demo keeps two of five chunks that cover all four claims, using 32% less of an 80-unit budget than retrieval-order packing. The evaluator works on structured observations and never reads the text or calls an LLM judge.

Observations persist in a local context graph backed by SQLite or PostgreSQL. A Python SDK exposes `prepare_context()`, and the MCP server is installed locally with `omneum install`. Supported platforms are macOS (Apple Silicon and Intel) and Ubuntu 22.04+ on x86_64, with Python 3.11 to 3.14.

## Features

- **Provenance-aware scoring:** Accounts for sources that share the same upstream origin.
- **Budgeted context selection:** Chooses passages by claim coverage within a token limit.
- **Text-blind evaluation:** Scores structured observations without reading text or using an LLM judge.
- **Persistent context graph:** Keeps observations across sessions in SQLite or PostgreSQL.
- **Python SDK:** prepare_context() returns the context, chosen chunks and evaluation scores.

## Good for

- Trim retrieval results before sending them to a model
- Prefer claims backed by independent sources over repeated ones
- Keep agent memory consistent across sessions
- Reduce prompt size without dropping key claims

## Install

### Claude Code (Terminal)

```sh
claude mcp add omneum -- npx -y https://github.com/cole-h06/omneum/releases/download/v1.0.4/omneum-local-1.0.4.mcpb
```

### Claude Desktop (claude_desktop_config.json)

```json
{
  "mcpServers": {
    "omneum": {
      "command": "npx",
      "args": [
        "-y",
        "https://github.com/cole-h06/omneum/releases/download/v1.0.4/omneum-local-1.0.4.mcpb"
      ]
    }
  }
}
```

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

```json
{
  "mcpServers": {
    "omneum": {
      "command": "npx",
      "args": [
        "-y",
        "https://github.com/cole-h06/omneum/releases/download/v1.0.4/omneum-local-1.0.4.mcpb"
      ]
    }
  }
}
```

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

```json
{
  "servers": {
    "omneum": {
      "type": "stdio",
      "command": "npx",
      "args": [
        "-y",
        "https://github.com/cole-h06/omneum/releases/download/v1.0.4/omneum-local-1.0.4.mcpb"
      ]
    }
  }
}
```

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

```json
{
  "mcpServers": {
    "omneum": {
      "command": "npx",
      "args": [
        "-y",
        "https://github.com/cole-h06/omneum/releases/download/v1.0.4/omneum-local-1.0.4.mcpb"
      ]
    }
  }
}
```

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

```json
{
  "mcpServers": {
    "omneum": {
      "command": "npx",
      "args": [
        "-y",
        "https://github.com/cole-h06/omneum/releases/download/v1.0.4/omneum-local-1.0.4.mcpb"
      ]
    }
  }
}
```

## FAQ

### Does it use an LLM to judge sources?

No, the README says evaluation uses structured observations and no LLM judge.

### Where is data stored?

Locally by default, in SQLite or PostgreSQL.

### Which platforms are supported?

macOS on Apple Silicon and Intel, and Ubuntu 22.04+ on x86_64.

### How do I install it?

pip install omneum for the SDK, then run omneum install without sudo to set up the local evaluator.

### Is Omneum free?

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

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

Run this command in your terminal: claude mcp add omneum -- npx -y https://github.com/cole-h06/omneum/releases/download/v1.0.4/omneum-local-1.0.4.mcpb

### What are the best alternatives to Omneum?

The closest alternatives to Omneum 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/omneum
