openaaas-mcp-adapter

# OpenAaaS MCP Adapter Connect Claude Desktop, Cursor, Cline and any MCP-compatible AI client to the [OpenAaaS Server](https://www.open-aaas.com). ## Quick Install (No Install Required) The recommended way to run the adapter — **no local installation needed**. Just make sure you have [uv](https://docs.astral.sh/uv/) installed: ```bash uvx openaaas-mcp-adapter ``` `uv` will automatically create a temporary virtual environment, fetch the package from PyPI, and run it. Nothing is left on your system. ## Client Configuration ### Claude Desktop Edit `~/Library/Application Support/Claude/claude_desktop_config.json` (macOS) or the equivalent path on your platform: ```json { "mcpServers": { "openaaas": { "command": "uvx", "args": ["openaaas-mcp-adapter"] } } } ``` ### Cursor In Cursor Settings → MCP, add a new server with the following configuration: ```json { "mcpServers": { "openaaas": { "command": "uvx", "args": ["openaaas-mcp-adapter"] } } } ``` ### Codex Add the adapter with the Codex CLI: ```bash codex mcp add openaaas -- uvx openaaas-mcp-adapter ``` For long-running task polling, configure `~/.codex/config.toml` (or `.codex/config.toml` in a trusted project): ```toml [mcp_servers.openaaas] command = "uvx" args = ["openaaas-mcp-adapter"] startup_timeout_sec = 30 tool_timeout_sec = 600 ``` Keep `poll_task(timeout_seconds=...)` below `tool_timeout_sec` so Codex has enough time to receive the tool result. Restart the Codex app, CLI session, or IDE extension after changing the configuration. ## Standard Workflow Once the adapter is connected, ask your AI agent to follow these steps: 1. **`set_server_url`** — Set the OpenAaaS server address (e.g. `https://api.open-aaas.com`) 2. **`register`** — Register a client name and obtain an API key (only needed once) 3. **`list_services`** — Browse available agent services 4. **`get_service_usage`** — Read the detailed usage guide for a chosen service 5. **`submit…