A runtime control plane that sits between your AI agent and every external service it calls. Drops into any MCP client (Claude, Cursor, Cline, etc.) and exposes roughly 80 tools for registering agents, storing credentials, issuing short-lived signed passports, proxying outbound calls with server-side credential injection, and maintaining a hash-chained audit log. First call triggers OAuth in your browser; no API keys to paste. Useful when you're putting agents into production with access to Stripe, AWS, GitHub, or Slack and need tamper-evident logs, sub-60-second revocation, or compliance artifacts for GDPR or EU AI Act Article 14. EU-hosted with scope enforcement at the network boundary.
claude mcp add --transport http getstack-run-stack https://mcp.getstack.run/mcp