01 / YOUR AGENT
Brings the current plan.
Provider, plan series, buyer tier and workload requirements. It can ask for monthly plans or another charging basis.
AgentCosts is the agent doing the cost check. Other agents call it through MCP; people use this website. Both run the same code against the same read-only market data.
Provider, plan series, buyer tier and workload requirements. It can ask for monthly plans or another charging basis.
Finds direct graph alternatives, selects relevant plans and returns their prices, actual histories and comparison flags.
Checks features and capacity, gets approval for a switch, and optionally reports whether it worked.
Agentery supplies the registry, observed plans, histories and accepted-substitute graph. Customers do not need a dataset. AgentCosts connects that evidence to a specific buying decision.
A Pro label is a useful starting point. Compare the captured billing terms and feature text before concluding that two plans can serve the same workload.
No. It provides evidence and watches. Your workflow must validate the fit and authorise any purchase or change through a separate integration — no AgentCosts tool can purchase, cancel or switch anything.
Yes. It is a live MCP tool: the reference plan series plus up to 8 alternatives, each with its own daily observations, like-for-like pair qualification, gaps and provenance flags. The website's evidence pages call the same service.
Yes. Watches are durable database records evaluated every day after the pricing refresh completes (not in real time). Events are deterministic and deduplicated; a disappeared plan prompts review and is never reported as a price change.
report_outcome is live and optional. Reports are correlated only when this endpoint actually served you evidence about that provider in the previous 30 days; reported costs are recorded as reported and never presented as verified savings.