# forecall-cli MCP server

> Scores how well MCP server tools are described for AI agents using an open-source linter and CLI.

- Repository: https://github.com/forecall/forecall-cli
- Page: https://hostedmcp.org/servers/forecall/forecall-cli
- Transport: stdio
- Stars: 0
- License: Apache-2.0

## Install

### Claude Code

```sh
claude mcp add forecall-cli -- npx -y forecall
```

### Cursor (`~/.cursor/mcp.json`)

```json
{
  "mcpServers": {
    "forecall-cli": {
      "command": "npx",
      "args": [
        "-y",
        "forecall"
      ]
    }
  }
}
```

### VS Code (`.vscode/mcp.json`)

```json
{
  "servers": {
    "forecall-cli": {
      "type": "stdio",
      "command": "npx",
      "args": [
        "-y",
        "forecall"
      ]
    }
  }
}
```

### VS Code CLI

```sh
code --add-mcp '{"name":"forecall-cli","type":"stdio","command":"npx","args":["-y","forecall"]}'
```

### Claude Desktop (`claude_desktop_config.json`)

```json
{
  "mcpServers": {
    "forecall-cli": {
      "command": "npx",
      "args": [
        "-y",
        "forecall"
      ]
    }
  }
}
```

### Codex

```sh
codex mcp add forecall-cli -- npx -y forecall
```

### Codex config.toml (`~/.codex/config.toml`)

```toml
[mcp_servers.forecall-cli]
command = "npx"
args = ["-y", "forecall"]
```

### Gemini CLI

```sh
gemini mcp add -s user forecall-cli npx -- -y forecall
```

### Gemini settings.json (`~/.gemini/settings.json`)

```json
{
  "mcpServers": {
    "forecall-cli": {
      "command": "npx",
      "args": [
        "-y",
        "forecall"
      ]
    }
  }
}
```

## FAQ

### How can I score my MCP server tools?

You can use the command `npx forecall lint tools.json` to run the linter on your tools definition file.

### Does the tool send my data to an external server?

No, the linter is designed so that nothing leaves your machine.

### Can I use this in a CI/CD pipeline?

Yes, you can use the Forecall GitHub Action to score your tools on every pull request and set quality gates.

### What metrics are used to score a tool?

Each tool is scored out of 100 based on purpose, usage context, arguments, return values, constraints/side effects, and examples.