webhook-toolkit
Allows receiving, forwarding, replaying, signing, and verifying GitHub webhooks, and sending signed test payloads to a local handler.
Allows receiving and forwarding Shopify webhooks to a local endpoint so handlers can be tested against live Shopify deliveries.
Allows receiving, forwarding, replaying, signing, and verifying Stripe webhooks, including generating realistic signed test events and diagnosing invalid signature errors.
Allows signing and verifying Twilio webhook requests, including diagnosing common signature mismatches such as URL, query string, or parameter ordering issues.
Click on "Deploy Server".
Wait a few minutes for the server to deploy. Once ready, it will show a "Started" state.
In the chat, type
@followed by the MCP server name and your instructions, e.g., "@webhook-toolkitcreate a webhook URL and forward events to localhost:3000/webhooks"
That's it! The server will respond to your query, and you can continue using it as needed.
Here is a step-by-step guide with screenshots.
webhook-toolkit
Receive, forward, sign and verify webhooks from your terminal, your test suite and your AI agent.
One package: a CLI (webhook-toolkit, alias whtk), a typed library and an MCP server, backed by webhook-toolkit.com.
Quick start
npx webhook-toolkit listen --forward http://localhost:3000/webhooksThat's it: no account, no config. You get a public URL; paste it into Stripe, GitHub, Shopify or any other sender, and every webhook shows up in your terminal and is re-sent to your local handler:
webhook-toolkit · listening
Webhook URL https://webhook-toolkit.com/r/l90c9MKIdo0o
Inspector https://webhook-toolkit.com/e/l90c9MKIdo0o
Forwarding → http://localhost:3000/webhooks
Expires in 7 days (2026-09-25 15:42)
Anonymous URL. Run `webhook-toolkit login` to get a permanent one.
Waiting for webhooks. Ctrl+C to stop.
15:42:11 POST /stripe stripe checkout.session.completed 248 B
↳ 200 OK · 3 ms
15:42:11 POST /github github push 269 B
↳ 500 Internal Server Error · 1 ms {"error":"handler crashed: cannot read properties of undefined"}Every request is also kept in a web inspector (headers, raw body, replay), so nothing is lost when your handler crashes.
Related MCP server: hookray-mcp
Contents
CLI
npm i -g webhook-toolkit # or run any command with npx webhook-toolkit …Command | What it does |
| Public URL + live stream of incoming requests, optionally re-sent to localhost. Reuses your last URL while it is alive ( |
| Real tunnel: callers get your local app's response. Paid. |
| Build a validly signed webhook: prints headers + a ready-to-run |
| Check a signature, and when it fails, say why. |
| Re-send a captured request from your machine (localhost works). |
| List captured requests ( |
| List your account's URLs. |
| Save, remove or inspect your API key. |
| Start the MCP server on stdio. |
A few things worth knowing:
Sub-paths and query strings are kept. A request to
…/r/<token>/stripe?x=1is forwarded tohttp://localhost:3000/webhooks/stripe?x=1, so one URL can feed several handlers.Headers are forwarded as received (minus hop-by-hop headers,
hostandcontent-length), so signature checks in your handler pass exactly as in production.--jsonprints NDJSON (listening,request,forwardevents) for scripting:whtk listen -f 3000 --json | jq .request.event.Colors follow
NO_COLOR,FORCE_COLORand are off when output is not a terminal.
Send a signed test event
# Prints the headers and a curl command, with a realistic sample event
whtk sign stripe --secret whsec_… --event checkout.session.completed
# Or send it straight to your handler
whtk sign github --secret my-secret --file push.json --send http://localhost:3000/api/github
whtk sign twilio --secret <auth token> --url https://example.com/sms --send 3000Debug "invalid signature"
whtk verify stripe --secret whsec_test_secret --body-file body.json \
-H "Stripe-Signature: t=1726000000,v1=a530ce3d81556a0eced506e1d1cad777bc55bc1e408e54dd6bfef08e1dde216c"✗ Invalid Stripe signature (body_trailing_newline_added)
The signature matches the body without its trailing newline: a newline was appended after signing (proxy, logger, copy-paste). Verify the raw bytes exactly as received.
expected t=1726000000,v1=1e0085bf1054f89ac8882e2e5fcbfdf301b1d6445e8ed7a316120cd5dc651ba9
received t=1726000000,v1=a530ce3d81556a0eced506e1d1cad777bc55bc1e408e54dd6bfef08e1dde216c
timestamp 1726000000 (738 days ago)The verifier replays the usual mistakes (whitespace in the secret, added or stripped trailing newline, CRLF conversion, expired timestamp, and for Twilio: http vs https, port, query string, trailing slash, repeated parameters) and tells you which one matches. The same logic powers the online signature validator.
Test webhooks in your test suite
The library gives your tests a real public URL and a way to wait for what lands on it. It works anonymously; set WEBHOOK_TOOLKIT_KEY in CI for permanent URLs and higher limits.
Vitest (or Jest in ESM mode)
import { expect, test } from "vitest";
import { WebhookToolkit, verify } from "webhook-toolkit";
const wt = new WebhookToolkit(); // anonymous, or reads WEBHOOK_TOOLKIT_KEY
test("creating an order notifies the customer's webhook", async () => {
const endpoint = await wt.createEndpoint({ name: "ci-orders" });
// The app under test sends its outgoing webhooks to the capture URL.
await api.post("/settings/webhooks", { url: endpoint.url, secret: "whsec_MfKQ9r8GKYqrTwjUPD8ILPZIo2LaLaSw" });
await api.post("/orders", { sku: "tshirt", quantity: 2 });
const req = await wt.waitForRequest(endpoint.token, {
timeoutMs: 30_000,
filter: (r) => JSON.parse(r.body).type === "order.created",
});
expect(JSON.parse(req.body).data.quantity).toBe(2);
// Your outgoing signatures are correct, too:
expect(verify("svix", { secret: "whsec_MfKQ9r8GKYqrTwjUPD8ILPZIo2LaLaSw", rawBody: req.body, headers: req.headers }).valid).toBe(true);
});waitForRequest checks requests that already arrived since the endpoint was created, then long-polls until timeoutMs, so the order of "trigger" and "wait" does not matter. It rejects with a WebhookToolkitError (code: "timeout") when nothing matched.
To test your handler without any network at all, sign the payload locally:
import { sign } from "webhook-toolkit";
const { headers, body } = sign("stripe", { secret: process.env.STRIPE_WEBHOOK_SECRET!, event: "checkout.session.completed" });
const res = await fetch("http://localhost:3000/api/webhooks/stripe", { method: "POST", headers, body });
expect(res.status).toBe(200);Playwright
import { expect, test } from "@playwright/test";
import { WebhookToolkit } from "webhook-toolkit";
const wt = new WebhookToolkit();
test("'Send test event' reaches the configured URL", async ({ page }) => {
const endpoint = await wt.createEndpoint();
await page.goto("/settings/integrations");
await page.getByLabel("Webhook URL").fill(endpoint.url);
await page.getByRole("button", { name: "Save" }).click();
await page.getByRole("button", { name: "Send test event" }).click();
const req = await wt.waitForRequest(endpoint.token, { timeoutMs: 15_000 });
expect(req.method).toBe("POST");
expect(req.headers["content-type"]).toContain("application/json");
});The package is ESM-only. CommonJS test runners work on Node ≥ 20.19 / 22.12 (native
require(esm)); Jest needs its ESM mode.
Use it from AI agents (MCP)
webhook-toolkit mcp is a Model Context Protocol server. Your coding agent can create a webhook URL, wait for the delivery after triggering it, read the payload, then replay it or send freshly signed events to your localhost handler while it fixes the code.
The API key is optional: everything except list_webhook_urls and explain_webhook_request works anonymously.
Claude Code
claude mcp add webhook-toolkit -- npx -y webhook-toolkit mcp
# with a key:
claude mcp add webhook-toolkit --env WEBHOOK_TOOLKIT_KEY=whk_… -- npx -y webhook-toolkit mcpCursor (.cursor/mcp.json), Windsurf (~/.codeium/windsurf/mcp_config.json), Claude Desktop (claude_desktop_config.json)
{
"mcpServers": {
"webhook-toolkit": {
"command": "npx",
"args": ["-y", "webhook-toolkit", "mcp"],
"env": { "WEBHOOK_TOOLKIT_KEY": "whk_…" }
}
}
}VS Code (.vscode/mcp.json)
{
"inputs": [
{ "type": "promptString", "id": "webhook-toolkit-key", "description": "webhook-toolkit.com API key (optional)", "password": true }
],
"servers": {
"webhook-toolkit": {
"type": "stdio",
"command": "npx",
"args": ["-y", "webhook-toolkit", "mcp"],
"env": { "WEBHOOK_TOOLKIT_KEY": "${input:webhook-toolkit-key}" }
}
}
}Codex CLI (~/.codex/config.toml)
[mcp_servers.webhook-toolkit]
command = "npx"
args = ["-y", "webhook-toolkit", "mcp"]
env = { WEBHOOK_TOOLKIT_KEY = "whk_…" }ChatGPT, Claude.ai and other remote-capable clients: add a custom connector pointing at the hosted server, nothing to install:
https://webhook-toolkit.com/mcp(Streamable HTTP, optional Authorization: Bearer whk_….) The hosted server cannot reach your machine, so replaying to localhost and send_signed_webhook need the local server above.
Tool | Use it to |
| Get a public URL to receive a webhook from a third-party service or the app under test |
| Block until the delivery arrives (filters: provider, event, method, path, body); safe to call after triggering |
| See what was actually sent: headers, raw body, detected provider and event |
| Re-send a captured request from the local machine to e.g. |
| POST a validly signed event (Stripe, GitHub, Shopify, Slack, Twilio, Mailgun, Svix, Paddle, Discord) to your handler |
| Compute signature headers and a |
| Explain why a handler rejects a signature |
| Change what the URL answers (status, body): simulate failures, answer challenges |
| Find an existing permanent URL (key required) |
| AI explanation of a payload, or a generated handler (paid, 3 free trials) |
Outputs are compact text plus the key JSON; bodies over 20 KB are truncated with a note pointing to the inspector. To embed the server in your own process: import { createMcpServer } from "webhook-toolkit/mcp".
Signing and verifying
sign() and verify() run locally with node:crypto: secrets never leave your machine. Signatures are byte-for-byte identical to the webhook-toolkit.com signer, and the tests check them against the providers' published examples (GitHub, Slack, Twilio, Svix).
Provider | id | Signature | Algorithm |
|
Stripe |
|
| HMAC-SHA256 hex of | endpoint secret |
GitHub, Gitea, Forgejo |
|
| HMAC-SHA256 hex of the body | webhook secret |
Shopify |
|
| HMAC-SHA256 base64 of the body | app client secret |
Slack |
|
| HMAC-SHA256 hex of | signing secret |
Twilio |
|
| HMAC-SHA1 base64 of URL + sorted params ( | auth token, plus |
Mailgun |
|
| HMAC-SHA256 hex of | webhook signing key |
Svix / Standard Webhooks (Clerk, Resend) |
|
| HMAC-SHA256 base64 of |
|
Paddle Billing |
|
| HMAC-SHA256 hex of |
|
Discord interactions |
|
| Ed25519 | verify: app public key; sign: a test private key |
import { sign, verify } from "webhook-toolkit";
// In an Express handler (use express.raw so the body stays byte-exact)
app.post("/webhooks/stripe", express.raw({ type: "*/*" }), (req, res) => {
const result = verify("stripe", { secret: process.env.STRIPE_WEBHOOK_SECRET!, rawBody: req.body, headers: req.headers });
if (!result.valid) return res.status(400).send(result.message);
// …
});verify() returns { valid, reason?, message, expected?, received?, timestamp? }. Failure reasons: missing_signature, malformed_signature, missing_timestamp, missing_url, invalid_secret, timestamp_out_of_tolerance (default 300 s, toleranceSeconds: 0 disables), signature_mismatch, secret_whitespace, body_trailing_newline_added, body_trailing_newline_removed, body_line_endings_changed, url_mismatch (Twilio), body_hash_mismatch (Twilio JSON). Comparisons are constant-time.
Provider detection
detectProvider(headers, body) returns { provider, event } from a data-driven table (DETECTION_RULES), for example { provider: "stripe", event: "invoice.paid" } or { provider: "twilio", event: "message.received" }. It knows Stripe, GitHub, GitLab, Bitbucket, Shopify, WooCommerce, Slack, SendGrid, Twilio, Paddle, Svix (Clerk, Resend), GoCardless, PayPal, Discord, Linear, Typeform, Calendly, Square, HubSpot, Zoom, Lemon Squeezy, Coinbase Commerce, Twitch, Mux, Vercel, Netlify, Sentry, Trello, Plaid, Mandrill, Jira, Postmark, Mollie, Adyen, Mailgun and Redsys. Adding one is a table entry plus a test fixture.
Relay: a real tunnel (paid)
listen --forward is free and fine for most webhooks, but the sender receives the capture URL's response, not your app's. When the sender needs your real answer (Slack slash commands, Twilio TwiML, Shopify mandatory responses, OAuth callbacks), use the relay:
whtk login # once
whtk relay --to 3000 webhook-toolkit relay · live
Public URL https://webhook-toolkit.com/relay/my-app (any sub-path works)
Forwarding → http://localhost:3000Requests to the public URL are proxied over a WebSocket to your machine and your local response (status, headers, body) goes back to the caller. It reconnects with backoff. The relay comes with the 7-day Pass or Pro.
Free vs paid
Anonymous | Free account | Pass (7 days, €5 once) / Pro (€9/month) | |
Capture URLs | up to 30 new per day, each lasts 7 days | 1 permanent URL | 25 permanent URLs |
History | 7 days | 7 days | 30 days |
CLI | ✓ | ✓ | ✓ |
Relay (real tunnel to localhost) | ✓ | ||
AI explanations and generated handlers | 3 free trials | 3 free trials | ✓ |
Details and the Business plan: webhook-toolkit.com/pricing. Everything that runs locally (signing, verifying, detection, forwarding) is free and stays free.
How it compares
smee.io is a free, open-source service from the GitHub Probot team that forwards webhooks to localhost through its client. If forwarding is all you need, it does the job. webhook-toolkit's
listen --forwardworks the same way and adds a stored inspector, provider/event detection, replay, signing and verification tools, a test-suite API and an MCP server.webhook.site is a mature hosted request inspector with custom responses and paid automation features. webhook-toolkit covers the inspect-and-respond part and focuses on the local developer loop: forwarding, signed test events, signature diagnosis, tests and agents.
ngrok is a general-purpose tunnel for any HTTP or TCP service, with a local traffic inspector. Pick it to expose a whole app. webhook-toolkit's free flow needs no account, and its paid relay is scoped to webhooks.
Library reference
import { WebhookToolkit, WebhookToolkitError, sign, verify, detectProvider, forwardRequest } from "webhook-toolkit";
const wt = new WebhookToolkit({ apiKey, baseUrl }); // both optional (env: WEBHOOK_TOOLKIT_KEY, WEBHOOK_TOOLKIT_URL)Method | Returns |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| server-side replay to a public URL |
| AI explanation or handler (paid, 3 free trials) |
| account, plan, relays |
forwardRequest(request, target) re-sends a captured request from your machine (the same function behind listen --forward and replay). Errors are WebhookToolkitError with status (HTTP status, 0 for network errors), code (unauthorized, upgrade_required, plan_limit, expired, timeout, network_error, …) and upgradeUrl on 402 answers. Everything is typed; types ship with the package.
Configuration
Variable | Purpose |
| API key ( |
| API origin, e.g. |
| Relay token for |
| Terminal colors. |
whtk login stores the key in ~/.config/webhook-toolkit/config.json (or $XDG_CONFIG_HOME/webhook-toolkit/config.json) with mode 600. The same file remembers the last URL used by listen.
A capture URL's token is a capability: anyone holding it can read the requests sent to it. Use an account for anything sensitive; the owner's key is then required to change or delete the URL.
Requirements
Node.js 18.17 or later. Runtime dependencies: @modelcontextprotocol/sdk and zod (MCP server) and ws (relay).
Contributing
Issues and pull requests are welcome: see CONTRIBUTING.md. New providers for detection or signing are the easiest place to start.
License
MIT © 2026 Kipdev. Hosted service: webhook-toolkit.com.
This server cannot be deployed
Maintenance
Related MCP Connectors
Debug webhooks from your AI agent: inspect and replay captured webhooks on localhost.
- webhook.coOAuthco.webhook
Receive, inspect, replay and deliver webhooks — with signature verification and agent triggers.
Webhooks for AI agents: send events, manage endpoints, inspect and retry deliveries.
Anonymous webhook capture, inspection, waiting, and response configuration for AI agents.
Related MCP Servers
- AlicenseNot gradedqualityDmaintenanceEnables generating webhook endpoints for testing, inspecting and comparing HTTP request payloads, replaying requests from history, and forwarding requests to localhost.2MIT
- AlicenseAqualityDmaintenanceEnables AI agents to create disposable webhook URLs, capture incoming HTTP requests, inspect headers and bodies, and replay them against local or remote endpoints, streamlining the webhook handler development loop.54 npmMIT
- AlicenseAqualityDmaintenanceWebhook management and testing tools for AI agents. Provides tools for sending, validating, generating, and debugging webhooks.525 npmMIT
- AlicenseAqualityCmaintenanceEnables AI agents to create callback endpoints, wait for async webhook results, and verify signatures, eliminating the need for polling.826 npm2MIT