Every call, logged and explained
Each search, read and run your agents make: which operation, how long it took, which account it used, and in the API’s own words why it failed.
Logs
What agents did through MCPAPI. Only argument names are kept: values and responses are never stored.
| When | Tool | Call | Status | Time |
|---|---|---|---|---|
| execute_tool | stripe.PostRefundsas live, approved in the chat(charge, reason) | completed200 | 412 ms | |
| get_tool_details | stripe.PostRefunds | completed | 38 ms | |
| search_tools | “refund a payment” | completed | 521 ms | |
| execute_tool | stripe.GetChargesas live(customer, limit) | completed200 | 288 ms | |
| search_tools | “list a customer’s charges” | completed | 498 ms | |
| execute_tool | asana.createTaskas default(name, assignee, due_on, projects) | completed201 | 655 ms | |
| execute_tool | polar.checkouts:create(products)This key is read-only, and polar.checkouts:create is a write. | denied | 12 ms | |
| execute_tool | linear.create_issue(title, teamId) | completed | 731 ms | |
| execute_tool | stripe.PostCustomersCustomeras sandbox(customer, email)Stripe: Invalid email address: ana@acme | error400 | 301 ms |
- Approvals on record
- A write that asked first shows that it was approved, and where.
- The whole trail
- Searches with their query, details, runs with their arguments’ names; never their values or your keys.
- Why it failed
- Refusals give the rule that stopped them; API errors quote the API.
Try it on your own APIs