# jaspertvdm/mcp-server-ollama-bridge [Health: Active]

**Category:** 🔗 Aggregators  
**Repository:** https://github.com/jaspertvdm/mcp-server-ollama-bridge  
**GitHub Stars:** 4  
**Views:** 3  
**Installs:** 0  
**Upvotes:** 0  
**Directory Page:** https://allmcps.com/mcp/jaspertvdm-mcp-server-ollama-bridge

## Description
Bridge to local Ollama LLM server. Run Llama, Mistral, Qwen and other local models through MCP.

## Claude Desktop Quick Installation
Install path detected from listing signals. Uses `uvx` (confidence: high):

```json
"mcpServers": {
  "mcp-server-ollama-bridge": {
    "command": "uvx",
    "args": ["mcp-server-ollama-bridge"],
    "env": {
      "OLLAMA_HOST": ""
    }
  }
}
```

**Requires environment variables:** `OLLAMA_HOST` — the values above are empty placeholders; fill in real credentials before running (see the repository for what each one is for).

## Documentation

## What jaspertvdm/mcp-server-ollama-bridge MCP server does

The jaspertvdm/mcp-server-ollama-bridge MCP server exposes a local Ollama installation to MCP clients. It is intended for running models hosted by Ollama through an MCP-compatible application rather than connecting directly to a remote model provider.

The bridge supports Ollama models generally, including the Llama, Mistral, and Qwen model families mentioned in the project description. The README also lists streaming responses as a supported feature, so clients can receive generated output incrementally when the Ollama server provides it.

## How it works

The MCP server acts as the connection layer between an MCP client and the Ollama HTTP service. By default, it looks for Ollama at `http://localhost:11434`. The target can be changed with the `OLLAMA_HOST` environment variable, which accepts the URL of the Ollama server.

For a local setup, the bridge process runs alongside the MCP client while Ollama runs on the host machine. The README shows Claude Desktop launching the `mcp-server-ollama-bridge` executable and passing `OLLAMA_HOST` in the process environment. Docker is also supported through the repository's Docker image, with the container configured to reach a host Ollama instance.

## Setup and configuration

Install the package from PyPI with:

```bash
pip install mcp-server-ollama-bridge
```

After installation, configure the MCP client to start `mcp-server-ollama-bridge`. Set `OLLAMA_HOST` to the Ollama URL, or omit it to use the default `http://localhost:11434`.

A Docker-based setup can build the repository image and run it interactively. The documented Docker example sets `OLLAMA_HOST` to `http://host.docker.internal:11434`, which is intended to let the container connect to Ollama on the host system.

## Tools and capabilities

The jaspertvdm/mcp-server-ollama-bridge MCP server provides these documented capabilities:

- Connect MCP clients to a local Ollama LLM server.
- Use Ollama models through MCP.
- Support all Ollama models, subject to what the connected Ollama installation makes available.
- Stream model responses.
- Select the Ollama endpoint through `OLLAMA_HOST`.

The provided material does not list individual MCP tool names, parameters, or model-management operations. It should therefore be evaluated as an Ollama access bridge rather than assumed to provide administrative controls for downloading, deleting, or inspecting models.

## Limitations and notes

The bridge depends on a separately running Ollama server. Installing the MCP package alone does not provide a model runtime or model files. The default endpoint assumes Ollama is available locally, while Docker deployments need network configuration that allows the container to reach the host service.

No API key or other credential is documented for this local connection. The README identifies Claude Desktop as a client configuration example; compatibility with other MCP clients is not specified in the provided material. The project is released under the MIT license.

_Full upstream README: https://allmcps.com/mcp/jaspertvdm-mcp-server-ollama-bridge/readme_