AI AGENT STORE / SERVICES

DURABLE WORKFLOWS & SCHEDULING · FOR AI AGENTS

Hookdeck Event Gateway

Queue, inspect and retry incoming webhooks that trigger agent workflows.

Hookdeck Event Gateway receives events before they reach an agent application, then routes and delivers them with controlled throughput. Its API, CLI and MCP tools let agents inspect deliveries, diagnose failures and manage connections.

Your integration checklist

What you need from your owner
Create an account and project, connect a source to a destination, and configure incoming webhook signature checks and destination authentication. Select throughput and retention appropriate to the workflow.
Authentication
Use a project API key for REST or authenticate the CLI with login or HOOKDECK_API_KEY. The beta MCP server runs locally over stdio with hookdeck gateway mcp; it has no remote server URL.
Your first task
Deliver a test webhook to a staging worker, make that worker fail temporarily, then inspect the retry and verify that the task is applied once.
Billing & payment
The Developer plan needs no credit card and includes limited monthly events and retention. Paid plans add usage and capacity. One inbound request can create multiple billable destination events; retries are included in event pricing.
Agent payment support
The limited Developer plan requires no credit card. Paid capacity, longer retention and overage arrangements require a separate billing decision.
Operational limits
Retention bounds recovery. MCP starts with limited write tools, but pause and unpause remain available without --allow-write. Excessive free-plan overages can stop processing; a queue is not an unlimited storage archive.

ORIGINAL AI AGENT STORE INTEGRATION PLAYBOOK

Make event retries safe before connecting them to agent actions

Editorial recommendation · Reviewed 2026-11-08

When to choose this service

Choose Hookdeck for durable inbound delivery and operational visibility across webhook providers. An existing reliable queue may already cover a simple internal producer.

Hosted ingress, buffering and delivery continue while an agent worker is offline. Writing a handler does not itself provide continuously operated event infrastructure or retained delivery history.

Suggested workflow

  1. Connect a staging source and destination with verified authentication. Record a stable provider event identifier.
  2. Make the consumer idempotent around that identifier. Test a timeout after the action succeeds so a retry cannot create a second action.
  3. Inspect failed attempts and queue age. Set delivery throughput to the worker capacity and plan recovery within the retention window.

Acceptance checks before real work

  • A successful webhook response is not proof that downstream work finished.
  • One request can fan out to several destination events.
  • The default MCP mode still allows operational pause and unpause actions.

If the workflow fails

If a backlog grows, check destination failures and worker capacity before replaying. Reconcile already-processed event identifiers, then retry only the outstanding work within retained history.

Provider documentation

Explore services for adjacent needs

Continue building your toolkit