Product
The product
A small set of processes, each with one job, and a console where your team configures them.
Ingest
A stateless endpoint that accepts the Segment-style HTTP tracking protocol. It checks the source’s write key, validates the batch, applies rate limits and answers only after the events are durably stored.
Router and workers
The router turns each event into one delivery job per enabled connection, after dedupe. Workers claim jobs, call the destination’s adapter and record the outcome: delivered, retrying, failed or dead.
Archive
The archiver writes every accepted event to S3-API object storage, which is what replays and backfills read from.
The console
A web app where members manage their workspace: sources and their write keys, destinations and their settings, connections and their rules, members and roles, and the audit log. SDK setup snippets for each source are on its page.
Built to be operated
Every process exposes Prometheus metrics, and the repository ships alert rules, a Grafana dashboard and runbooks for the failure modes it knows about: a backlog, an open circuit, dead letters, a rate-limited source.
Limits at a glance
- Request size
- 512 KiB
- Events per batch
- 1,000
- Default rate per source
- 1,000 events/s
- Default rate per workspace
- 5,000 events/s
- Delivery retries
- Up to 25 attempts or 24 hours
- Write key rotation grace
- 24 hours by default
Send once, deliver everywhere
Sign in, create a source, connect your destinations and change one host setting in your SDK.