webhook.co vs Hookdeck

Why people compare these two

When people go looking, it tends to be the invoice — and specifically its shape rather than its size. Hookdeck's Event Gateway meters delivered events, but it also meters throughput as a separate purchase — the default ceiling is five events per second per destination — and it meters discarded requests, which are the requests your own filters threw away. Both paid tiers are advertised as starting at a price rather than being one. The second reason is quieter and only affects some teams: the read-only role and the directory-sync features start at the four-hundred-and-ninety-nine-dollar tier.

The interesting difference here is not a feature, it is how many numbers you need to predict a bill. Hookdeck's Event Gateway prices on a plan fee, a tiered rate for delivered events, a separately purchased throughput ceiling in events per second per destination, a metered charge for discarded requests beyond an included allowance, a retention window fixed by tier, and a static-IP add-on. Every one of those is a defensible engineering decision and several are genuinely fairer than a flat rate at scale. But six dials is six things to model before you know what next month costs. We meter one thing — captured events — and when a free organization runs out, capture pauses rather than billing you. That is a deliberate constraint on us as much as a feature for you: it means we cannot invent a second meter later without breaking our own rule.

At a glance

Where Hookdeck is straightforwardly ahead

We are pre-launch and they are not, and most of this list follows from that. They publish a component-level status page and contractual uptime and latency commitments on their upper tiers. We publish a status page too, but make no contractual uptime commitment — our own terms disclaim one. They have a first-party Terraform provider, so a platform team can manage connections as code. They have the compliance paperwork that gets a vendor through procurement, and a trust centre to serve it from; we hold no certifications and say so on our privacy page. Their issue model is more developed than ours — delivery, transformation and backpressure issues as first-class objects, with a configurable queue-time threshold and two-way sync into incident tooling. They support arbitrary JavaScript transformations on every tier including free. And their outbound product is Apache-2.0 and genuinely self-hostable, delivering to queues and brokers rather than only to HTTP.

How many numbers are in the bill

Worth walking through concretely, because the throughput dial is the one that surprises people. A Hookdeck project includes five events per second of delivery throughput per destination; above that, events are queued rather than rejected, which is the right behaviour, but their latency commitment explicitly does not apply to delivery above your purchased throughput. So a burst-heavy source does not cost you events, it costs you either latency or a throughput purchase. Separately, requests your filters discard count against an allowance, and past it they are billed per million — you are paying a little for traffic you configured the system to throw away. Our meter is the captured request, disclosed at the point you create an endpoint, with a soft cap that pauses rather than surprises.

Provider coverage is roughly a tie

We would rather say this than let a table imply otherwise. Their documentation lists 146 source types, of which 141 are named third-party platforms, and their public OpenAPI specification defines provider-specific verification configurations for 135 of them. Our registry covers 141 providers. Those numbers are close enough that provider coverage should not decide anything between us, and anyone telling you otherwise is counting selectively. The real difference is that ours is Apache-2.0 and readable, and that we ship an in-browser verifier you can paste a captured request into without an account.

What an agent is allowed to do

Both of us have shipped MCP servers, and they are scoped very differently. Hookdeck's runs locally over stdio through their CLI, and the MCP server itself is read-only by design: their documentation is explicit that it exposes no tools to create or update sources, destinations, connections or transformations, and that retrying an event is not available through it. That is not the whole picture, and their own page says so — writes are routed through their published Agent Skills instead, so a Hookdeck agent can create and change things, just not through the MCP surface. Ours is a remote server with OAuth where the writes live in the MCP itself, and that includes destructive ones: an agent holding our token can delete an endpoint, delete captured events and rotate an ingest URL. We withhold delivery destinations and payload bodies deliberately, because an agent that can redirect where your events go is a confused-deputy problem — but we should be straight that our surface is the more powerful and therefore the more dangerous of the two, and that theirs is a defensible way to draw the line.

When to choose Hookdeck

If you are putting this in front of production traffic this quarter, choose Hookdeck. They have the operational maturity, the incident tooling, the contractual commitments and the compliance evidence, and we have a pre-launch product and a promise. Choose them if you need to fan out to queues and brokers rather than HTTP endpoints, if you need infrastructure-as-code, if you need transformations you can write in JavaScript, if you need to satisfy a security questionnaire, or if you are at a volume where their per-event rate card falls well below our flat tiers. Choose them if your organisation needs a read-only role. That is a lot of reasons, and they are all real.

When to choose us

If a predictable single-meter bill matters more than a rate card that optimises at scale. If you want the receiving path itself to be open source rather than the outbound half. If you want your agents to act on events over a remote MCP server rather than only read them. And if replaying a captured event straight to a port on your laptop is part of your daily loop rather than an occasional debugging move.

Moving from Hookdeck

Sources, and how to correct us

Every claim above about Hookdeck was read off the pages below on the date shown, not from memory. Products change faster than pages do — if something here is wrong or out of date, including if you work at Hookdeck, tell us and we'll fix it.

• Hookdeck pricing — checked 2026-07-22
• Hookdeck public OpenAPI specification (verification configs) — checked 2026-07-22
• Hookdeck docs — source types — checked 2026-07-22
• Hookdeck Outpost pricing and licence — checked 2026-07-22
• Hookdeck docs — MCP server — checked 2026-07-22
• Hookdeck TypeScript SDK repository — checked 2026-07-22

Other comparisons

• webhook.co vs RequestBin
• webhook.co vs ngrok
• All comparisons