A webhook pipeline your agent can drive — keyless. Provision a real endpoint, wait for the provider event, replay it into localhost, forward it to your app, and generate types from your actual captured payloads. Built for humans debugging webhooks and for AI agents building the integration.
→ App: https://webhook-studio.com · Docs: https://webhook-studio.com/docs · Agent spec (llms.txt): https://webhook-studio.com/llms.txt
- Debug — a zero-signup endpoint, a live inbox with full headers + raw body, and signature verification results on every request.
- Route — forward matching events to your backend with AND/OR conditions, and durable re-drive: a delivery interrupted by a crash or deploy is re-driven from a durable record, not silently dropped.
- Learn — every event updates a durable model of what the provider actually sends your account. Ask for a TypeScript / Zod / JSON-Schema type and the optional fields are the ones genuinely sometimes absent in your traffic; ask what changed between two versions and get a breaking-change diff. No other webhook tool does this. Honest comparison: https://webhook-studio.com/compare
Webhook Studio is a remote MCP server (Streamable HTTP, stateless). Add one URL:
claude mcp add --transport http webhook-studio https://webhook-studio.com/mcp \
--header "Authorization: Bearer $(curl -s -X POST https://webhook-studio.com/api/start | jq -r .api_token)"That curl mints a keyless read + replay token (no signup) bound to a fresh 3-day endpoint. For
full access — create endpoints, configure verification, stand up forwards — use a ws_live_… key
from your dashboard.
Or add it to your client's MCP config (~/.cursor/mcp.json, .vscode/mcp.json,
claude_desktop_config.json) — full setup for every client at https://webhook-studio.com/mcp/setup:
{
"mcpServers": {
"webhook-studio": {
"url": "https://webhook-studio.com/mcp",
"headers": { "Authorization": "Bearer ws_live_your_key_here" }
}
}
}The Authorization header is optional — omit it for a public connection (the read/replay tools that
accept a ws_anon_… token), or set a ws_live_… key for the full toolset.
Listed in the official MCP Registry as com.webhook-studio/webhooks.
The manifest is server.json in this repo.
Task-shaped, not resource-shaped — reach for them by intent:
create_bucket · list_buckets · get_bucket · delete_bucket |
Provision and manage endpoints |
latest_event · list_events · get_event · wait_for_event |
Read events; block until one arrives |
replay_event · forward_bucket · list_deliveries |
Replay to localhost, forward, check results |
configure_verification |
Turn on HMAC with the provider's real secret |
list_schemas · get_schema · diff_schema · generate_types · find_correlation_keys |
The knowledge layer |
Each tool maps 1:1 onto a /v1 REST route, so the REST reference
is also the tool reference.
curl -X POST https://webhook-studio.com/api/start
# → { ingest_url, inbox_url, api_token: "ws_anon_…", expires_in_days: 3, ... }Point a provider at ingest_url, hand the human inbox_url, and use api_token to read + replay
over /v1 or /mcp. Full agent guide: https://webhook-studio.com/llms.txt
One-click test URLs and correct signature-verification snippets per provider:
Documentation and examples in this repo are MIT. The hosted product is proprietary.