moa-mcp

A local stdio MCP server (mixture of agents) for running single-turn prompts across multiple model providers - OpenAI, Ollama (cloud models proxied locally) and OpenRouter. Similar in spirit to Claude's advisor tool, but not limited to one family of models: agents can fan the same prompt out to several models and combine the output.

Tools

• list_models - returns supported providers and their models, each with a human-friendly name and the provider-specific id to pass to execute_prompt.

• execute_prompt - executes a single-turn prompt against a given provider and model_id. Optional attachments is a list of local file paths (text or images) included as context; paths work because the server always runs on the same machine as the caller.

Related MCP server: AgentTasker MCP Server

Setup

cp .env.example .env   # then fill in the keys
uv sync

Ollama must be running locally (ollama serve) and signed in to ollama.com for the cloud models.

Registering with Claude Code

claude mcp add --scope user moa -- uv run --directory /path/to/moa-mcp moa-mcp

Development

uv run pytest

Providers live in src/moa_mcp/providers/, one module each, all delegating to a shared OpenAI-compatible chat helper in base.py. To add a model, extend the hardcoded list in the relevant provider module.