API Orchestration
Call, chain, and retry any HTTP API with durable state, typed inputs, and full traces.
Durable calls between systems
API Orchestration lets a workflow call any HTTP API and carry the result into the next step. Each call is made from a durable runtime, so a worker restart or a downstream outage does not lose the run. The workflow resumes from the step that failed.
Calls are configured in code with typed inputs and outputs. A developer defines the request, the expected response, and the retry policy, and the platform enforces them on every execution.
What engineers get
- Retry policies with backoff, jitter, and a maximum attempt count for each call.
- Rate limiting per destination, so a burst of events cannot overwhelm a partner API.
- Fan-out and fan-in, so a workflow can call several APIs in parallel and combine the results.
- Idempotency keys that prevent a retried call from creating a duplicate record.
Traces for every run
Every run produces a trace that lists each call with its request, response, timing, and retries. Engineers can search traces by workflow, status, or payload field, and they can replay a run against a new version of the code.
Traces can be exported to the logging and monitoring tools your team already uses, so orchestration shows up in the same dashboards as the rest of your services.
Teams that call a dozen partner APIs from one workflow rely on traces to find the single slow dependency, rather than guessing which service is responsible.