Local AI MCP runs on your own machine — the client starts it, so there is no endpoint to ping. 24 installs a week from npm. Last commit 3 Nov 2026.
Run, fit, and route work to local LLMs (Ollama, LM Studio, llama.cpp, MLX, vLLM, and more).
Today is the operative word: we check Local AI MCP every 15 minutes and re-read its code on every release. Watch it and you find out the day that stops being true.
This server runs on your own machine — install it with the package manager and the client starts it for you. Package name taken from the official registry entry.
claude mcp add local-ai-mcp -- npx -y @tmhs/local-ai-mcp
{
"mcpServers": {
"local-ai-mcp": {
"command": "npx",
"args": [
"-y",
"@tmhs/local-ai-mcp"
]
}
}
}
[mcp_servers.local-ai-mcp]
command = "npx"
args = ["-y", "@tmhs/local-ai-mcp"]
{
"mcpServers": {
"local-ai-mcp": {
"command": "npx",
"args": [
"-y",
"@tmhs/local-ai-mcp"
]
}
}
}
{
"servers": {
"local-ai-mcp": {
"type": "stdio",
"command": "npx",
"args": [
"-y",
"@tmhs/local-ai-mcp"
]
}
}
}
Optional settings from the registry entry, left out of the block above:
OLLAMA_HOST (Ollama base URL), LMSTUDIO_HOST (LM Studio server base URL), LLAMACPP_HOST (llama.cpp server (llama-server) base URL), DOCKER_MODEL_RUNNER_HOST (Docker Model Runner host TCP endpoint), MLX_HOST (mlx_lm.server base URL (opt-in; Apple Silicon)), VLLM_HOST (vLLM OpenAI server base URL (opt-in)), VLLM_API_KEY (Bearer token for VLLM_HOST (when vLLM runs with --api-key)), FOUNDRY_LOCAL_HOST (Foundry Local service URL (opt-in; port from `foundry service status`)), OPENAI_COMPAT_HOST (Any OpenAI-compatible /v1 server (opt-in)), OPENAI_COMPAT_API_KEY (Bearer token for OPENAI_COMPAT_HOST), MOONSHOT_API_KEY (Moonshot AI (Kimi) key. Hosted: only used when named explicitly), LOCAL_AI_ALLOW_HOSTED_FALLBACK (Allow implicit routing to hosted providers (default: off)), MOONSHOT_HOST (Moonshot API base URL), LOCAL_AI_DISABLE (Comma-separated provider names to skip probing), LOCAL_AI_PROVIDERS (JSON array of extra provider instances: {type, host, name?, apiKey?, kind?}), LOCAL_AI_DISABLE_NETWORK_SEARCH (Keep search_models offline (catalog only)), LOCAL_AI_BENCHMARK_CACHE (Benchmark results file used to rank delegate by measured speed (default: per-user cache dir; off disables)).
QSP — relay MCP tools to any OpenAI-compatible local LLM (llama.cpp, Ollama, vLLM)
Bring Visual Studio-style C# fixes and refactorings to your AI agent with local Roslyn Code Actions.
Let a planning agent hand bounded language work to a local Ollama model.
MCP server that offloads routine coding tasks to local LLM (Ollama) to save Claude API tokens
Run Wiro AI models, build and run workflows, and talk to Studio and your Wiro agents.
Fan a goal out to cheap parallel LLM workers on your own key or a free local Ollama. Keep your mind.
Connect Claude to any OpenAI-compatible LLM endpoint and offload routine work to a local model.
Offload cheap work from your AI agent to a local Ollama model, at zero cloud cost.
Answers built from our own checks of this server.