Use RunAPI in Cursor or Claude Code to discover AI models, inspect inputs and current pricing, check your account balance, and create media tasks.
The plugin connects to https://mcp.runapi.ai/mcp through browser-based OAuth.
- Obtain this plugin directory from the RunAPI plugin repository.
- Copy the complete directory, including its dotfiles, to
~/.cursor/plugins/local/runapi. - Run Developer: Reload Window in Cursor.
- Open Customize and confirm that the RunAPI skill and MCP server appear.
Your Cursor administrator must allow local plugin imports.
From the downloaded plugin directory, start Claude Code with:
claude --plugin-dir .The plugin loads for that session.
- Open the client's MCP connection settings and authenticate the
runapiserver. - Complete RunAPI sign-in and approve the requested access in your browser.
- Ask the agent to call RunAPI's
list_modelstool. Confirm that it returns model identifiers and capabilities.
The client stores the OAuth session. The plugin requires no API key or environment variables. If the session expires, reconnect through the client's MCP settings. Model discovery verifies the connection without creating a paid task.
The plugin is free and open source under the Apache License 2.0. RunAPI service usage can require paid credits. Check current pricing through the MCP tools or the model catalog before authorizing a task. The included skill asks for explicit approval before paid actions.
Authorized requests send their prompts, input URLs, and task parameters to RunAPI. Account and task results return through your MCP client. Review the privacy policy and terms of service before connecting. Download generated files into your own storage before their URLs expire.
Run these checks from the plugin directory before distributing a modified copy:
python3 scripts/validate.py
claude plugin validate . --strictThe Python validator uses only the standard library. It checks both manifests, the OAuth MCP configuration, the skill, the license, and the logo.