Migrate from Hookdeck
Use HookSense for agent callbacks, Hookdeck for app event routing.
Hookdeck is a great event gateway for routing events between applications. But it has no wait_for_callback primitive an AI agent can block on. HookSense is the agent-native callback layer — your agent creates an endpoint and awaits the signed result over MCP.
Step 1 · Identify what to migrate
- 1In Hookdeck, find the flows where an AI agent waits on a single result — long tool calls, human approvals, external async jobs.
- 2Decide which are agent-callback flows vs. which need app-to-app queueing/SLA delivery.
- 3Agent-callback flows move to HookSense. App-to-app event routing stays on Hookdeck.
Step 2 · Map features
| In Hookdeck | In HookSense |
|---|---|
| Source (e.g., `stripe`) | Agent-created callback endpoint via create_callback_endpoint |
| Consumer polls the queue for events | Agent calls wait_for_callback and blocks until the signed result arrives |
| Manual signature wiring | verify_signature built into the MCP toolset |
| Issue tracker (failed deliveries) | Auto-retry and dead-letter — coming soon (paid plans in early access) |
| 1 source → N destinations fan-out | Not 1:N delivery — HookSense is the single-result agent callback layer |
Step 3 · Things to know
- If you use Hookdeck for 1-source → N-destinations app-to-app fan-out, HookSense is NOT a replacement. Pair the two: Hookdeck routes events, HookSense handles agent callbacks.
- Auto-retries and dead-letter are coming soon and available in early access on paid plans — same intent as Hookdeck's retry, scoped to the agent callback.
- HookSense is MCP-native: install with `npx @hooksense/mcp` so your agent owns the callback endpoint lifecycle.
Zero-downtime migration
Install `npx @hooksense/mcp` and migrate one agent-callback flow at a time. Keep Hookdeck for app-to-app event routing; HookSense for the agent callback loop. Paid plans in early access.
Get a callback URL