Build and manage your chatbot from outside the dashboard. The Agent Gateway exposes your flows, knowledge base, storage, and conversations as tools, callable by an AI agent over MCP or by your own code over a REST API.
What the gateway is, what an agent or a script can do with it, and how access is scoped and kept safe.
Create a token, choose its scopes, pin it to workspaces, and revoke it when you are done.
Point Claude Code, Claude Desktop, Cursor, or a custom agent at your workspace over MCP.
Call the same tools over plain HTTP from a script or backend that does not speak MCP.
Every tool the gateway serves, with the scope it needs and the arguments it takes. Generated from the running gateway, so it always matches what your agent can call.
Create, edit, publish, and deploy flows through the same tools you use by hand, and run them in test mode to check the work before it goes live.
Add and reprocess knowledge sources, and manage your storage tables and their rows, so what the bot answers from stays current.
List conversations and read transcripts, and with the right scope inspect a run's step trace, to see how the bot is doing.
The gateway is not only for AI agents. Drive the same tools from a script or backend over REST to automate management tasks on a schedule.