MCP Server
QuetraAI provides an MCP (Model Context Protocol) server that exposes governance tools to any MCP-compatible AI agent. Zero code required — just add a server URL to your config.
Two Connection Options
| Method | Setup | Multi-agent | Requires Node.js |
|---|---|---|---|
| Remote (recommended) | Paste one URL | Yes | No |
| Local (stdio) | npx @quetra/mcp | No | Yes |
Remote Server URL
https://mcp.quetra.dev/<your-api-key>/mcp
Replace <your-api-key> with your API key from the dashboard. The remote server supports multi-agent routing — use quetra_list_agents to discover agents.
Available Tools (9)
Governance tools evaluate, track, and report on spending. Connector tools execute authenticated actions on third-party services via Composio and Arcade.dev. Quetra governs first; connectors execute second.
| Tool | Purpose |
|---|---|
| quetra_list_agents | Discover agents with budget details and rule summaries (remote only) |
| quetra_evaluate | Governance check — approves/rejects spending, tracks budget |
| quetra_check_budget | View remaining budget and limits |
| quetra_can_spend | Quick approval check with rejection reasons |
| quetra_fetch | x402-aware fetch for paid APIs |
| quetra_transactions | View transaction history |
| quetra_execute | Execute an authenticated action via Composio or Arcade.dev (connector) |
| quetra_connect_service | Get auth URL for connecting a third-party service (connector) |
| quetra_list_provider_tools | Discover available tools for a service/toolkit (connector) |
Platform Guides
Claude Desktop / Code
Step-by-step Cowork integration with remote and local options.
Setup guideOpenClaw
Connect via remote MCP server with openclaw.json config.
Setup guidenpm Packages
External Resources
- MCP Protocol Specification — official Model Context Protocol docs
- Anthropic MCP Guide — building with MCP in Claude