Skip to main content
Glama
LukeLamb

claude-ollama-mcp

by LukeLamb

Claude Ollama

Lets Claude Desktop query and manage a local Ollama server. List installed models, inspect them, run one-shot generate/chat completions against any local model, or pull/delete models from the registry — all without opening a terminal.

Typical use: comparing Claude's answer to a local model on the same prompt, running cheap bulk completions against a quantized model, or checking custom training-checkpoint models you've imported into Ollama.

Requirements

  • A running Ollama server (ollama serve or the Ollama app).

  • Default endpoint is http://localhost:11434. Override via the ollama_url user config in Claude Desktop's extension settings if you run Ollama on a different host or port.

  • No npm dependencies — pure Node over the HTTP API.

Related MCP server: Ollama MCP Server

Install (Claude Desktop)

  1. Download the latest Ollama.mcpb from the Releases page.

  2. In Claude Desktop: Settings → Extensions → Extension Developer → Install Extension → pick the .mcpb.

  3. (Optional) In the extension's settings, set Ollama server URL if you run Ollama on a non-default host/port. Leave blank for http://localhost:11434.

Tools

Tool

Annotation

Purpose

ollama_status

read-only

Health check + server version

list_models

read-only

Local models with size, digest, family, parameter size, quantization

list_running

read-only

Models currently loaded in VRAM

show_model

read-only

Model details: modelfile, parameters, template, capabilities

generate

open-world

One-shot text completion (non-streaming)

chat

open-world

Chat completion with message history (non-streaming)

pull_model

open-world

Download a model from the registry

delete_model

destructive

Remove a locally-installed model

Example prompts

"Which local models do I have installed, and which one is currently loaded in VRAM?"

"Run forge:b6c1 on this prompt: ''. Compare that output to your own answer."

"Show me the modelfile for forge:b7c1 — I want to check the temperature setting."

"Pull llama3.1:70b." (expect a long wait for large models)

"Delete the forge:b5c3 model — I don't need that checkpoint anymore."

Privacy policy

This extension runs entirely on your local machine and sends HTTP requests only to your Ollama server (default http://localhost:11434). No data leaves your machine unless you explicitly configure ollama_url to point at a remote Ollama instance, in which case the prompts and responses travel to that server.

The information visible to Claude includes:

  • All prompts and chat messages you pass to generate and chat (these go to the Ollama server, which may log them depending on its configuration).

  • Full text of completions returned by Ollama.

  • Metadata for every installed model (names, digests, sizes, quantization, modelfile contents).

  • Which models are currently loaded in VRAM and their size footprint.

If you have installed models containing proprietary fine-tunes or modelfiles with sensitive metadata, note that Claude will see that information when you call show_model or list_models.

delete_model is destructive and cannot be undone from this extension — the model must be re-pulled from the registry (or re-imported from source blobs) if deleted by mistake.

Troubleshooting

"cannot reach Ollama at http://localhost:11434 — is the server running?" — Start Ollama with ollama serve or launch the Ollama app. Verify with curl http://localhost:11434/ (should return "Ollama is running").

pull_model hangs for a long time — Ollama's pull API with stream: false blocks until the full download completes, which for multi-GB models can take many minutes. If you're pulling a huge model, run ollama pull <name> in a terminal instead — you'll see streaming progress there, and subsequent MCP calls will find the model already installed.

Custom/remote Ollama endpoint — Set ollama_url in the extension's settings (e.g. http://192.168.1.42:11434). Requires restart of the extension.

list_running shows a model after you stopped using it — Ollama keeps models hot in VRAM for a configurable TTL (default 5 minutes). The expires_at timestamp tells you when it'll unload. This is Ollama's behavior, not the extension's.

Development

Single ~400-line Node.js script, zero npm dependencies. Rebuild the .mcpb:

cd bundle-source
zip -j ../Ollama.mcpb manifest.json package.json server.js README.md LICENSE icon.png glama.json

License

MIT. See LICENSE.

Install Server
A
license - permissive license
A
quality
C
maintenance

Maintenance

Maintainers
Response time
Release cycle
1Releases (12mo)
Commit activity

Resources

Unclaimed servers have limited discoverability.

Looking for Admin?

If you are the server author, to access and configure the admin panel.

Related MCP Servers

  • A
    license
    -
    quality
    C
    maintenance
    Enables interaction with locally running Ollama models through chat, generation, and model management operations. Supports listing, downloading, and deleting models while maintaining conversation history for interactive sessions.
    714
    MIT
  • A
    license
    -
    quality
    D
    maintenance
    A bridge that enables Claude Code to interact with local Ollama instances for text generation, multi-turn chat, and vision-based analysis. It supports model management tasks such as listing, pulling, and showing details, alongside generating text embeddings.
    714
    MIT
  • A
    license
    -
    quality
    D
    maintenance
    Exposes local Ollama instances as tools for Claude Code, allowing users to offload code generation, text drafting, and embedding tasks to local GPUs. It supports multi-turn conversations and model management through the Model Context Protocol.
    MIT

View all related MCP servers

Related MCP Connectors

  • Operate Linux, macOS and Windows from your LLM. Every action runs through an auditable allowlist.

  • Live SEO workflow tools for Claude Code, Codex, and AI agents.

  • WHOOP recovery, strain, sleep and workouts in Claude via official WHOOP OAuth. Free, open source.

View all MCP Connectors

Latest Blog Posts

MCP directory API

We provide all the information about MCP servers via our MCP API.

curl -X GET 'https://glama.ai/api/mcp/v1/servers/LukeLamb/claude-ollama-mcp'

If you have feedback or need assistance with the MCP directory API, please join our Discord server