Skip to main content

Webhook Data Capture for MCP Servers

Some MCP servers accept push data from upstream systems (exports, form submissions, scheduled jobs) through a dedicated ingest URL. That flow is different from Webhook Triggers, which start A2A agent runs.

Webhook ingest (this page)Webhook triggers
PurposeStore incoming payloads for tools on an MCP serverWake an A2A agent with event context
URL lives onConnection panel for a webhook ingest providerAgent Configuration → Webhook Trigger
Typical useCSV/JSON drops, partner webhooks, cron exportsZapier → agent run, Stripe → agent

When you see it​

On an MCP server Connections row, connect a provider that uses the webhook ingest panel. After you save the connection, the panel shows:

  • Ingest URL — HTTPS endpoint upstream systems POST to (treat as a secret)
  • Last delivery — timestamp when MCPBundles last received a payload
  • Webhook auth — URL token only, or HMAC via X-MCPBundles-Signature: sha256=<hex> on the raw body when a signing secret is set

Before the first save, the panel explains that you must save the connection to mint the ingest URL.

Setup flow​

  1. Open the MCP server → Connections.
  2. Click Connect on the webhook-capable provider.
  3. Name the connection if prompted, set an optional signing secret (when your sender supports custom headers), and Save.
  4. Copy the Ingest URL into the upstream system (Zapier webhook step, monitoring tool, cron script, form builder, etc.).
  5. Send a test POST; the panel should show Last delivery received.
  6. In your AI client, call the server’s export/read tools (names vary by listing) to pull the captured data into the conversation.

Security​

  • Rotate by creating a new connection or follow listing-specific docs if the provider supports token rotation.
  • Prefer HMAC when the sender can set headers; use URL-token-only mode only when the upstream cannot (some SaaS schedulers).