AI Web API Agent: Autonomous Testing for Modern APIs

Where the Web API Viewer gives operators a fast way to browse and manually launch tests against an API, the AI Web API Agent takes the next step: it plans and runs a testing pass largely on its own. Point it at an OpenAPI document, a HAR capture from a browser session, or a raw HTTP request, and it builds a working model of the target API from whatever format is on hand.

The agent sits on top of RTFMv2's shared AI backbone -- AiManager and its provider clients for OpenAI, Ollama, LM Studio, and llama.cpp, the same abstraction used by the framework's narrative generation and knowledge base features -- so it inherits whichever local or hosted model an operator has already configured, rather than requiring a separate API key or setup step.

From there, the agent is designed to reason about which endpoints are worth prioritizing and drive categories of testing -- SQL injection probes, parameter fuzzing, and authenticated request replay -- without requiring an operator to hand-configure each pass. For teams facing large, sprawling API surfaces under time pressure, it is meant to compress the setup-and-triage phase of API testing so more time goes toward validating and escalating real findings.