
[ Beta ] // v0.1
Our customers interact with our API on a daily basis to automate their revenue workflows. Our MCP server, integrates natively with agents by exposing a curated set of tools which wrap our API.
A customer adds the MCP URL in Claude, which authenticates through the OAuth 2.0 protocol. Once permissions have been granted, Claude can directly authenticate with our MCP endpoint with a Bearer token encoding its permissions. Streamable HTTP is used to facilitate low-latency, continuous dialogue between the MCP server and the client.
We’re incrementally adding more tools to enhance the capabilities of the agent using the MCP server. We launched with a set of read-only tools wrapping our GET API endpoints as a proof-of-concept, and we’re expanding into write tools. The security layer is paramount here, to ensure the agent has limited access to the actions it has been granted. We’re also building up a prompt library which composes tool calls to illustrate the use cases the MCP server can enable.
Labs gets sharper when the people who'd actually use these agents push back. Pick your way in.