Platform

The domain pages describe what IonWarp does. This page indexes the contract underneath all of them — the same tasks runtime, the same Data objects, the same timeline, the same money model, whichever door a request came through. Each entry says what you will find there; none of it is restated here.

The runtime

• Tasks — one unit of agent work: create it, start it, read its outcome, read its transcript. The lifecycle every skill executes inside.
• Debugging a Run — the status block field by field, success versus visible failure, and the named failure classes a caller can branch on.
• Control Plane — projects, run modes, the approval brake, and the spend cap: the operator controls that decide whether a run may start at all.

The objects

• Data — read and write objects in a project's collections: data_set, data_get, data_query, data_batch. One snake_case spelling on both MCP and REST.
• Timeline — the ordered, human-readable event stream on every object, with provenance and idempotent writes.
• API reference — every collection this agent serves, with its real fields and a cURL · AI Prompt · MCP call for each operation.

Capability and cost

• Skills — the live catalog, and how to enumerate and read IonWarp's executable skill bodies at run time (skills_list / skills_get) so a caller runs the live contract rather than a copy.
• Integrations — provider credentials, the credential-blind call proxy, the capability catalog, and the risk classes that decide what an agent may execute.
• Usage — credits, the per-run itemized cost, and what happens when a cap is reached.

Cross-cutting reference

• MCP Tools — every MCP tool with its real arguments, generated from the live registry.
• Error Codes — the REST error envelope { error: { code, message, request_id } } and each code's default HTTP status.
• Authentication — per-route, per-method auth mode and the PAT scope each one requires.
• Pagination — limit + offset, the bounded list envelope, and why has_more is the only signal that another page exists.
• Conventions — id prefixes, snake_case fields, epoch-millisecond time, and credit-denominated money.
• llms.txt — these docs as plain text, for an agent to read in one fetch.