# offline-mcp [Health: Active]

**Category:** 🗣️ Conversational AI  
**Repository:** https://github.com/gabrielmahia/offline-mcp  
**GitHub Stars:** 0  
**Views:** 1  
**Installs:** 0  
**Upvotes:** 0  
**Directory Page:** https://allmcps.com/mcp/offline-mcp

## Description
Local AI inference infrastructure for East Africa — Ollama, open weights, degraded-mode fallbacks. 6 tools. Thesis L4.

## Tools
Capabilities this server exposes over MCP:

- **check_ollama_status** — Check whether a local Ollama server is running at localhost:11434 and which models it has. Use before run_local_inference, or when someone asks whether offline AI is ready. Takes no arguments and contacts only your own machine (3 second timeout). Returns ollama_running, the model names, a count and the endpoint, or ollama_running false with the error and a link to the official download page. Read-only: it only lists models.
- **run_local_inference** — Send a prompt to a local Ollama model and return its answer, with no internet needed once the model is downloaded. Use when no cloud model is reachable or the text must stay on the machine. The prompt goes only to Ollama at localhost:11434 (60 second timeout, non-streaming) and nothing is stored by this tool. Returns the response text, the model used, the token count and offline true, or an error with a hint to start Ollama if it is not running. Read-only with respect to your data; it uses local compute, and answers from small models can be wrong.
- **list_recommended_models** — Recommend open-weights models that run locally through Ollama, with size, RAM need, use case and language fit for Africa. Use when someone asks which model to install for their hardware or task. max_ram_gb keeps models that fit; use_case filters by matching text and, if nothing matches, ALL models are returned. Returns the source label, the recommended models, their install commands and a note that they run locally. A curated list labelled by the author, not a benchmark: check model cards. Read-only; no side effects.
- **degraded_mode_guide** — Guide to running AI when connectivity or power is unreliable: why it matters, how to install Ollama, pull a small model, test it and point MCP clients at localhost, with hardware minimums and a Raspberry Pi note. Use when planning an offline or low-bandwidth deployment. Takes no arguments. Returns the explanation, numbered setup steps, hardware minimums and deployment tips. A static guide: follow the vendor's own install instructions at ollama.com. Read-only; no side effects.
- **open_weights_directory** — Directory of open-weights model families with sizes, languages and licences, including African-language support such as Aya and Masakhane work, plus speech and translation resources (WAXAL, Paza, AfriNLLB, TranslateGemma) with their licences. Use when someone asks which open models or datasets cover Swahili or other African languages, or what the licence allows. use_case filters the language models by matching text; if nothing matches, ALL models are returned. Returns the source label, the language models, an Africa note, how to deploy them locally and the speech and translation resources with licence and coverage caveats. A curated starting list, not exhaustive or benchmarked: check each model card for languages and licence. Read-only; no side effects.
- **local_deployment_guide** — Setup guide for running a local model on a device type: laptop, Raspberry Pi and others in the table, with hardware, install steps and a model suggestion. Use when someone asks how to deploy offline AI on specific hardware. An unrecognised device_type returns the laptop guide. Returns the source label, the device_type, the guide, all_options and a solar note. A static guide: follow the vendor's install instructions. Read-only; no side effects.

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

```json
"mcpServers": {
  "offline-mcp": {
    "command": "uvx",
    "args": ["offline-mcp"]
  }
}
```

## Documentation

## What offline-mcp does

The offline-mcp MCP server provides an MCP interface for local AI infrastructure built around Ollama. It combines two kinds of capability: operational access to a locally running inference service, and reference material for selecting models and planning offline deployments.

The server is aimed at environments where internet access is intermittent or unavailable. The README identifies Kenya, Tanzania, and Uganda as primary deployment contexts and describes use in clinics, schools, community offices, and other local settings. Its stated design keeps inference and community data on the device rather than sending prompts to a hosted model provider.

The package is also positioned as part of an East Africa coordination stack.

## How it works

The operational path uses Ollama running locally. `check_ollama_status` reports whether Ollama is available and returns the models it can access. `run_local_inference` sends a prompt to a selected local Ollama model. The README names Llama 3.2 and Qwen 2.5 as examples of models that can run on the device, but the available choices depend on what has been installed locally.

The offline-mcp MCP server follows a local-first approach: core operations are intended to work without internet connectivity. Its reference material also describes graceful degradation, hardware-aware model selection, and a four-level degraded-mode architecture. A JSONL queue is listed among the design principles for accumulating events offline and synchronizing when connectivity returns, although no queue-specific MCP tool appears in the six-tool list.

The project describes a Raspberry Pi 4 deployment powered by a 50W solar panel and using a 256GB SD card as an example context. That example indicates the intended operating range, not a guarantee that every model or workload will run on that hardware.

## Setup and configuration

Install the Python package with:

```bash
pip install offline-mcp
```

Before using inference, an Ollama installation must be running locally and must have at least one usable model. The server documentation does not specify environment variables, additional configuration files, authentication settings, or a required model name.

Hardware and model selection remain deployment decisions; the included guides are intended to help compare laptop, server, Raspberry Pi, and Android options.

## Tools and capabilities

Six MCP tools are documented:

- `check_ollama_status`: Checks local Ollama availability and lists available models.
- `run_local_inference`: Runs a prompt through a local Ollama model.
- `list_recommended_models`: Recommends open-weight models for East Africa use cases.
- `degraded_mode_guide`: Explains a four-level architecture for operating with reduced connectivity or capability.
- `open_weights_directory`: Provides a directory of open-weight models with Africa language support.
- `local_deployment_guide`: Covers deployment considerations for laptops, servers, Raspberry Pi, and Android.

Together, these tools support both immediate local inference and planning work. For example, an agent can inspect the local Ollama installation, identify suitable models, and consult deployment guidance before running a prompt.

## Limitations and notes

The offline-mcp MCP server depends on Ollama for its inference tool, so local inference is unavailable if Ollama is not running or no suitable model is installed. Model performance and language coverage will depend on the selected open-weight model and the available hardware.

The README presents the project for offline and East African use cases, but it does not provide benchmarks, resource requirements for each model, or a complete hardware compatibility matrix. It also labels demo data as unsuitable for operational decisions. Treat the guides as deployment reference material and validate models, hardware, and outputs for the intended workload.

_Full upstream README: https://allmcps.com/mcp/offline-mcp/readme_