A concrete audit of where requests, model calls, database queries, mock traffic and license checks run in the Powerduck desktop app, and what the network can and cannot see.
Integration is done and the partner asks for a link. Finish descriptions locally, verify real calls, and publish versioned hosted docs and MCP from the same workspace, with no doc-tool rollout attached.
Inventory APIs get consumed by developers reading docs and by agents discovering tools. Separate web docs, prompt pastes, and hand-written schemas guarantee drift; one OpenAPI file produces both.
A concrete Cloud walkthrough: publish a managed Streamable HTTP MCP endpoint from one spec, expose twelve operations out of forty-seven, gate agents with a rotated bearer key, inject upstream auth, and pin the partner to an immutable version.
Hand-writing tool wrappers around existing order and inventory endpoints means maintaining every contract twice. Generate the MCP server from the OpenAPI spec, start with read-only tools, and widen the surface deliberately.
A real integration walkthrough: import the vendor OpenAPI file once, let the assistant and your coding agent read the exact contract, and send real requests locally, with no more curl archaeology in chat windows.
A grounded look at the review gates that keep AI-generated APIs usable: staged plans, granular JSON Patch cards that never overwrite your fields, read-before-edit tools, and deterministic validation.
The create call returned 200, the detail view 404s, and cancel accepts an illegal state. Scenario tests chain real calls, pass data between steps, and hand the agent evidence to fix with.
A concrete sprint diary: one OpenAPI file drove a local mock with latency and error overrides, a four-step checkout scenario with an HTML report, live docs, derived database tables and a hosted MCP endpoint while the real API did not exist yet.
Fake objects in components demo well and integrate badly. Agree on the coupon validation contract first, then let a spec-driven mock answer real HTTP while the service is still being built.
AI agents need tools, and hand-writing MCP wrappers around your API recreates the exact drift problem OpenAPI solved. Powerduck Cloud turns one spec into hosted docs and a ready-to-use MCP endpoint.
Generated screens fail on limit versus pageSize, data.records versus data.items, cents versus dollars. The fix is not a better prompt; it is letting the agent query the real contract and verify against real responses.
