What is replay?
Replay is re-running a retained event with the same input — typically after fixing a bug that caused the original to fail or after changing the consumer's logic.
The basic pattern is straightforward: pull a stored event, re-deliver it to the destination, observe the result. The hard part is making replay deterministic — re-running with the exact same input that produced the original failure.
Why deterministic replay is harder than it sounds
Three things have to be true for replay to reliably reproduce the original failure:
- The original payload bytes are persisted (not just the parsed payload — the bytes the sender actually sent)
- The transformation applied to the payload is persisted (not the current transform config — the snapshot of what was applied at original delivery time)
- The replay mechanism uses both (not the current config plus the persisted payload, which would produce a different result)
Most home-grown webhook systems get one or two of these right and miss the third. The result: replays produce different outputs than the original, and the fix-and-verify loop becomes guesswork.
Why agents need replay more than classical software
Classical webhook handlers are deterministic — same payload in, same response out. AI agents aren't — the LLM's output varies, tool calls have real-world side effects, retries cost tokens. Without deterministic replay, you can't reproduce a customer's failure to fix it.
Hooksbase persists raw payloads and dispatch snapshots per delivery while they are inside the payload retention window, and exposes one-click replay from the dashboard.
For the deeper read: Deterministic replay for agents.
Frequently asked questions
What makes replay deterministic?
Three things have to be true at once: the original payload bytes are persisted, the transformation applied at the time is persisted, and replay uses both rather than the stored payload plus the current config. Most home-grown systems get two of the three, and their replays then produce different output than the original.
Why do AI agents need replay more than classical handlers?
A classical handler is deterministic — same payload in, same response out. An agent is not: the model's output varies, tool calls have real-world side effects, and every retry costs tokens. Without deterministic replay you cannot reproduce a customer's failure in order to fix it.
What is Hooksbase?
Hooksbase is event infrastructure for AI agents. It ingests events over four channels — HTTP, email, HTML form, and scheduled cron — verifies them, routes them by rule, runs versioned Automations in the event path, and delivers them to HTTP and cloud destinations (AWS SQS, AWS EventBridge, GCP Pub/Sub, and S3-compatible storage) with retries, strict ordering, Standard Webhooks-compatible signing, deterministic replay, and a dead-letter path. It is a hosted service, runs on Cloudflare Workers, is operated at hooksbase.com, and is not affiliated with — and shares no code or ownership with — other similarly named webhook, hook, or tunnelling tools.
Related terms
- Dispatch snapshot
The transformed payload at the moment of dispatch, persisted so replays remain correct under config change.
Read - Idempotency
A property where performing an operation multiple times produces the same result as performing it once.
Read - DLQ (Dead-Letter Queue)
A holding area for messages that failed terminally so they can be inspected and recovered.
Read - Retry
Re-attempting a failed delivery on a backoff schedule until it succeeds or hits a retry budget.
Read