# Mneme Decision MCP

> Mneme Decision MCP is an open-source MCP server for developer tools. Local MCP server giving coding agents your architectural decisions, plus proposals. It runs locally on your machine and is started by your AI client.

- **Kind:** MCP server
- **Pricing:** Open source
- **Platforms:** CLI, self-hosted
- **Website:** https://mnemehq.com/integrations/mcp/
- **Repository:** https://github.com/mnemehq/mneme
- **MCP Registry name:** io.github.MnemeHQ/mneme
- **Categories:** developer tools, knowledge and memory
- **Listing updated:** 2026-11-14

## About

Agents read the repository's recorded decisions and can suggest new ones, but they cannot make a suggestion binding. Proposals are saved as non-authoritative records with source provenance and a stable ID, and a person accepts or rejects them through the separate `mneme decision` CLI. Authority-shaped input fields are rejected at the schema boundary.

The server exposes exactly six tools, unchanged since 0.9.0. Two create proposals, including `decision.propose` and `decision.propose_batch`, and four are read-only: search, contextual matching such as `decision.applicable_to`, and lineage tracing. It reads the Decision Index in `.mneme/project_memory.json` and refuses to start until that file has been migrated.

It runs only as a local stdio process, with no remote endpoint, cloud service or role-based access layer. Enforcement stays with Mneme's typed rules, hooks and CI gates, which are deterministic and do not use an LLM judge. The engine is MIT licensed with no paid tier.

## Features

- **Read-only decision queries:** Agents can search decisions, find contextual matches and trace known lineage.
- **Non-authoritative proposals:** Candidate decisions are stored as proposed, with provenance and a stable ID.
- **Human acceptance:** Only the mneme decision CLI can accept or reject a proposal.
- **Fail-closed startup:** The server will not run until project memory carries a valid Decision Index.
- **Frozen tool surface:** Six tools, deliberately small and unchanged since 0.9.0.
- **Local stdio transport:** Runs as a local process launched by the MCP host, with no hosted control plane.

## Good for

- Give a coding agent the decisions that apply to its current task
- Let an architecture tool submit candidate decisions for review
- Trace where a recorded decision came from
- Keep agent suggestions separate from enforced rules

## Install

### Claude Code (Terminal)

```sh
claude mcp add mneme -- uvx mneme-hq
```

### Claude Desktop (claude_desktop_config.json)

```json
{
  "mcpServers": {
    "mneme": {
      "command": "uvx",
      "args": [
        "mneme-hq"
      ]
    }
  }
}
```

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

```json
{
  "mcpServers": {
    "mneme": {
      "command": "uvx",
      "args": [
        "mneme-hq"
      ]
    }
  }
}
```

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

```json
{
  "servers": {
    "mneme": {
      "type": "stdio",
      "command": "uvx",
      "args": [
        "mneme-hq"
      ]
    }
  }
}
```

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

```json
{
  "mcpServers": {
    "mneme": {
      "command": "uvx",
      "args": [
        "mneme-hq"
      ]
    }
  }
}
```

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

```json
{
  "mcpServers": {
    "mneme": {
      "command": "uvx",
      "args": [
        "mneme-hq"
      ]
    }
  }
}
```

## Tools

- `decision.propose`: Creates a non-authoritative proposal record for a candidate decision.
- `decision.propose_batch`: Creates several non-authoritative proposal records at once.
- `decision.applicable_to`: Finds decisions that apply to a given task.

## FAQ

### Can an agent enforce or accept a decision?

No. MCP cannot accept or reject proposals, activate or supersede decisions, install rules or submit trusted evidence.

### How do I start it?

Install with pip install "mneme-hq[mcp]" and run mneme decision-mcp. Use --memory to point at another project memory file.

### Is there a remote or hosted endpoint?

No. This release is a local stdio server only, with no HTTP or SSE transport.

### What does it cost?

The engine is free under the MIT License and the pricing page says there is no paid tier today.

### Is Mneme Decision MCP free?

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

### How do I add Mneme Decision MCP to Claude Code?

Run this command in your terminal: claude mcp add mneme -- uvx mneme-hq

### What are the best alternatives to Mneme Decision MCP?

The closest alternatives to Mneme Decision MCP are REA, Codebase Memory, GitHub, Serena MCP: the IDE for your agent and XERJ.

## Alternatives

- [REA](https://ai-powered.com/mcp/rea): REA is an open-source MCP server for developer tools.
- [Codebase Memory](https://ai-powered.com/mcp/codebase-memory): Codebase Memory is an open-source MCP server for knowledge and memory.
- [GitHub](https://ai-powered.com/mcp/github-2): GitHub is an open-source MCP server for developer tools.
- [Serena MCP: the IDE for your agent](https://ai-powered.com/mcp/serena-mcp-the-ide-for-your-agent): Serena MCP: the IDE for your agent is an open-source MCP server for developer tools.
- [XERJ](https://ai-powered.com/mcp/xerj): XERJ is an open-source MCP server for knowledge and memory.
- [Ouroboros](https://ai-powered.com/mcp/ouroboros): Ouroboros is an open-source MCP server for developer tools.
- [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.

---
Source: https://ai-powered.com/mcp/mneme-decision-mcp
