I will test your ai agent tool calls apis and regressions

Your AI agent may look perfect in a demo and still call the wrong tool, send malformed arguments, repeat a side effect, or silently break after a prompt or model change.

I will build a focused, reproducible reliability test suite for an existing AI agent that uses APIs, functions, webhooks, databases, or MCP tools. This is engineering QA for agent behavior, not generic prompt feedback.

I can test:

correct tool selection and argument schemas

expected tool-call sequence and task completion

invalid inputs, timeouts, API failures, and retry behavior

duplicate calls and unsafe side effects in a sandbox

response grounding against returned tool data

regressions across prompt, model, or workflow changes

latency and call usage where available

You receive runnable Python tests or a structured test pack, captured failures with reproduction steps, and a prioritized report. Standard and Premium can include agreed fixes; every claimed fix is re-tested.

I work only with sandbox/test credentials and non-destructive scenarios. Please message me before ordering so I can confirm your agent and access method are testable.