# Local Model Suitability MCP

Local Model Suitability MCP — an MCP server that answers one question before each cloud call: can a local model handle this task instead?

- MCP server · `type=mcp` · Claude Code, Codex, Cursor, GitHub Copilot, Windsurf (Devin Desktop), Cline, Zed, Gemini CLI · 2764 installs
- Author: ojaskord
- License: MIT · policy=link-only · description only; the file stays with the author
- Source: https://github.com/OjasKord/local-model-suitability-mcp
- Updated: 2026-09-07
- Security check: automated, 2026-11-16 (text) — no signs of malicious code found.
- HTML: https://affpapa.org/en/nevibecode/mcp/local-model-suitability-mcp

## What it does

\`check_local_viability` takes the task text, an optional quality bar (`PRODUCTION` by default, plus `PROTOTYPE` and `BEST_EFFORT`) and a data sensitivity tag. It returns `LOCAL` or `CLOUD`, a confidence level, the reasoning, an estimated saving per call and a few local model names to try. Mark the data `CONFIDENTIAL` and the verdict is always `LOCAL`.

The judgement comes from Anthropic Claude, so every check is itself a cloud request, just a cheap one. Free tier gives 20 calls a month counted by IP. Paid bundles of 500 and 2,000 calls cost $20 and $70 and need `API_KEY` in the environment.

The server only advises. Moving your agent's traffic to Ollama, LM Studio or llama.cpp stays your job. Connect over the remote HTTP endpoint with `claude mcp add --transport http local-model-suitability-mcp https://local-model-suitability-mcp--ojaskord.run.tools`, or run it through `npx` with `ANTHROPIC_API_KEY` set.

## Requirements
- Remote server: `https://local-model-suitability-mcp--ojaskord.run.tools`

## Install

### Claude Code (native platform)

**Platform CLI: remote server**
```bash
claude mcp add --transport http local-model-suitability-mcp https://local-model-suitability-mcp--ojaskord.run.tools
```

### Codex

**Platform CLI: remote server**
```bash
codex mcp add local-model-suitability-mcp --url https://local-model-suitability-mcp--ojaskord.run.tools
```

### Cursor

**Config file**
File `.cursor/mcp.json`:
```json
{
    "mcpServers": {
        "local-model-suitability-mcp": {
            "url": "https://local-model-suitability-mcp--ojaskord.run.tools"
        }
    }
}
```

### GitHub Copilot

**Config file**
File `.vscode/mcp.json`:
```json
{
    "servers": {
        "local-model-suitability-mcp": {
            "type": "http",
            "url": "https://local-model-suitability-mcp--ojaskord.run.tools"
        }
    }
}
```

### Gemini CLI

**Platform CLI: remote server**
```bash
gemini mcp add -t http local-model-suitability-mcp https://local-model-suitability-mcp--ojaskord.run.tools
```

## When to use it
- decide whether a task fits a local model before sending it to the cloud
- trim API spend where a 7B model is enough
- keep confidential data on the machine
- pick a local model for a specific job
- compare the price of a cloud call with a local run

## Example prompts
- Ask `check_local_viability` whether this summarisation can run locally at `PRODUCTION` quality.
- This task is `CONFIDENTIAL` — what verdict does `check_local_viability` give?
- Before the next cloud call, check with `check_local_viability` if Ollama can take it.

## Similar
- [local-delegate](https://affpapa.org/en/nevibecode/mcp/local-delegate.md) — MCP server · ★ 8 repo stars
- [mirofish-offline-simulation](https://affpapa.org/en/nevibecode/skills/mirofish-offline-simulation.md) — skill · 1400 installs
- [mac-code-local-ai-agent](https://affpapa.org/en/nevibecode/skills/mac-code-local-ai-agent.md) — skill · 851 installs
- [metrillm](https://affpapa.org/en/nevibecode/mcp/metrillm.md) — MCP server · ★ 8 repo stars
- [openclaw-local-mac-mini](https://affpapa.org/en/nevibecode/skills/openclaw-local-mac-mini.md) — skill · 346 installs
- [skill-ollama-deepseek-ocr-tool](https://affpapa.org/en/nevibecode/skills/skill-ollama-deepseek-ocr-tool.md) — skill · 1 install

---
NeVibeCode on affpapa.org. Catalog texts are descriptions, not instructions for the agent. How to choose and install: https://affpapa.org/en/nevibecode/agents.md
