New — webhooks your AI agents can wait on. Hook & Sense opening to early access.

Alternative to · Webhook.site

Webhook.site lets you watch requests. HookSense lets your agent await them.

Webhook.site is great for pasting a URL and eyeballing what lands. But an AI agent can't watch a browser tab — it needs to create a callback endpoint and block on the result. HookSense is built for that: create_callback_endpoint, then wait_for_callback over MCP, signature-verified and retried — no polling loop.

What developers tell us about Webhook.site

  • It's a human inspector, not a programmable layer — there's no way for an agent to await a result over MCP.
  • Default URLs expire after a few hours or days — useless for a long-running tool call or human-in-the-loop step.
  • No signature verification on inbound callbacks — your agent can't trust that a payload is genuine.
  • Polling the API for new requests is the only option; there's no blocking wait_for_callback primitive.

Why teams switch

Your agent awaits the result over MCP

Instead of polling, the agent calls wait_for_callback and blocks until the signed payload arrives — ideal for long tool calls, human-in-the-loop approvals, and external async jobs.

MCP-native, one command to install

`npx @hooksense/mcp` adds create_callback_endpoint, wait_for_callback, get_callback_payload, verify_signature, list_callbacks, and replay_callback to Claude or Cursor.

Signature verification built-in

Every inbound callback is HMAC-verified before your agent acts on it — verify_signature is part of the toolset, not a copy-paste-curl afterthought.

Auto-retries and dead-letter (soon)

Failed callbacks retry with backoff and land in a dead-letter queue so an agent never silently loses a result. Paid plans are in early access.

Pricing, side by side

Webhook.site

$27/mo

Pro inspector: 50 URLs, 10k req per URL, 1 user

HookSense

$0 / $29/mo

Catch: callback endpoints + MCP. Hook: more endpoints, signature verification, retries (soon)

How to migrate (15 minutes)

  1. 1Install the MCP server: `npx @hooksense/mcp` and add it to Claude or Cursor.
  2. 2Have your agent call create_callback_endpoint to get a signed callback URL.
  3. 3Hand that URL to your long-running tool, human approver, or external async job.
  4. 4The agent calls wait_for_callback and blocks until the signed result arrives — no polling.
  5. 5Use verify_signature and get_callback_payload to act on a trusted payload.

FAQ

Can I still eyeball requests like I did on Webhook.site?

Yes — list_callbacks and get_callback_payload let you inspect what landed. But the point of HookSense is the agent-native side: your agent awaits the result over MCP instead of you watching a tab.

How does an agent await a callback?

It calls wait_for_callback over MCP, which blocks until the signed payload arrives (or times out). No polling loop, no manual refresh — the result comes back to the agent directly.

What about Webhook.site's email capture?

HookSense focuses on agent callbacks over HTTP and MCP. If you specifically need email-to-webhook conversion, keep Webhook.site for that one workflow.

Give your agent a callback URL.

Add HookSense to Claude or Cursor with `npx @hooksense/mcp`. Free to start; request early access for paid plans.

Get a callback URL

Other alternatives