Postel vs Svix vs Outpost vs Convoy vs DIY
An honest comparison across footprint, cost, transactionality, portal, and SLAs — including what Postel deliberately doesn't do.
The real question is "where does webhook delivery live." Svix and Hookdeck Outpost are dispatcher services you run (or rent) alongside your app. Convoy is the same shape, self-hosted. A hand-rolled BullMQ worker is the DIY version of that shape. Postel is a library: no dispatcher, no broker, your existing database.
Pick based on which shape fits, not which table has more checkmarks — see Is Postel for me? for the decision gate.
| Postel | Svix | Hookdeck Outpost | Convoy | DIY (BullMQ) | |
|---|---|---|---|---|---|
| Shape | Embedded library | Hosted service (self-host on Enterprise only) | Self-hosted service (Hookdeck-managed option) | Self-hosted service (Convoy Cloud option) | Hand-rolled worker |
| Footprint | None beyond your existing Postgres/MySQL/SQLite | A separate dispatcher; self-hosting gated behind an Enterprise plan | Separate Go service + Redis + Postgres + a queue (SQS/RabbitMQ/Kafka/…) | Separate Go service + Postgres + a Redis HA cluster/Sentinel | Redis (for BullMQ) + code you write and maintain |
| Cost | $0 — no service to run or pay for | Free tier (200 msg/sec), then $490/mo+ (Professional), custom Enterprise | Free self-hosted (you pay the infra); Hookdeck-managed tiers on top | Free self-hosted (MIT); paid add-ons (e.g. SSO ~$99/mo) | $0 license, but signing/rotation/replay/dedup are yours to build and maintain — the hidden cost is engineering time |
| Transactionality | Outbox row shares your business-write transaction — commit or rollback together | None — you call their API after your commit; a gap exists between your write and their ingestion | None — your app publishes an event to a downstream service; no shared transaction | None — ingested via HTTP API after the fact | None out of the box — enqueuing to Redis is a separate network hop from your DB commit unless you build the outbox pattern yourself |
| Customer-facing portal | Not offered (see below) | Yes — embeddable, themable, a headline feature | Yes — built-in end-user portal | Yes — subscriber portal | None — you'd build it |
| SLA | None — it's your process, your app's SLA is your SLA | Tiered: 99.9% (free/startup) → 99.99% (Professional) → 99.999% (Enterprise) | None self-hosted; Hookdeck's managed tier carries its own SLA | None self-hosted; Convoy Cloud carries its own SLA | None — self-managed |
What we deliberately don't do
Per VISION.md, Postel will never:
- Be a service or host one
- Run a separate dispatcher process
- Require Redis, RabbitMQ, Kafka, or any message broker
- Ship a customer-facing portal as a packaged product
- Compete on multi-region replication or five-nines SLAs
If any of those are what you actually need, Svix or Hookdeck Outpost are the right tools — not a gap Postel is trying to close.
Where the atomic-outbox difference actually matters
Every dispatcher-shaped option above (Svix, Outpost, Convoy, and DIY-BullMQ without extra work) has the same failure window: your business write commits, then a separate call tells the dispatcher about it. If the process dies between those two steps, the event is silently lost — no error, nothing to retry, because the dispatcher never heard about it.
Postel's send() runs inside the same database transaction as the business write. Kill the process mid-transaction and both roll back together; nothing half-commits. See the demo at the top of the README or run it yourself in examples/nextjs-prisma.
Read next
- Is Postel for me? — the six-line decision gate.
- Sending & the outbox — the transactional contract this page compares against.
- Migration guides — moving from BullMQ, Svix, or a signing-only library onto Postel.