Pricing
What it costs
Five offers. One of them is free, three are how most customers start, and the last is for teams who already have a connector and only want the testing.
| Directory-readiness scan | Build, test, file | Hosted connector | Pre-flight CI | |
|---|---|---|---|---|
| Price | Freeno account, no card | $2,500one-time | $79per month | $199per month |
| Who it is for | Anyone with a domain or an endpoint | A business with a public API | Every build customer | Teams who built their own connector |
| A readiness report per directory | Included | Not included | Not included | Not included |
| A connector from one definition: MCP and OpenAPI | Not included | Included | Not included | Not included |
| An idempotency key on every write | Not included | Included | Not included | Not included |
| A run against the fault-injection harness | Not included | Included | Not included | Included |
| A replayable trace and a failing test per finding | Not included | Included | Not included | Included |
| Submission to Claude, ChatGPT and Muse | Not included | Included | Not included | Not included |
| Owner approval before anything commits you | Not included | Not included | Included | Not included |
| Agent identity checks and attribution | Not included | Not included | Included | Not included |
| Re-filing when a directory changes its rules | Not included | Not included | Included | Not included |
| An API layer over your booking or quoting system | Not included | add $1,500 | Not included | Not included |
What is included
Scan — Free
- A one-page report per directory: Claude, ChatGPT, Muse.
- Every finding names the rule it fails and the change that fixes it.
- A shareable link you can send to whoever owns the API.
- No account, no card, nothing installed.
Build, test, file — $2,500 one-time
- A connector built from your API: an MCP endpoint for Claude and ChatGPT and an OpenAPI document for Muse, from one definition.
- Every write takes an idempotency key, so a retry cannot double-book.
- A pre-flight run against the fault-injection harness, with the trace and a failing test you keep.
- The listing text, screenshots and test cases each directory asks for, generated from your connector rather than written from memory.
- Submission to Claude, ChatGPT and Muse, and the follow-up until each one answers.
- Add $1,500 if we also build the API layer over your booking or quoting system.
Hosted connector — $79/month
- We run the connector on Cloudflare and keep it up.
- Owner approval on WhatsApp before anything spends money or books a slot.
- Identity checks on which agent is calling, using signed agent requests where the agent supports them.
- Attribution: which assistant asked, how often, and how many of those turned into work.
- Re-filing when a directory changes its rules.
Pre-flight CI — $199/month
- The harness on its own, pointed at a connector you already own.
- Duplicate calls, tool errors, expired sessions, injected instructions, two agents racing for the same slot.
- Deterministic checks first; a model is only asked the one question a machine cannot settle, which is whether the agent claimed a success that did not happen.
- A replayable trace and a failing test per finding, in your repository, not ours.
The usual ones
Questions
- Who owns the connector?
- You do. The definition, the generated code and the tests are yours, and you can take them and host them yourself whenever you like.
- What if a directory rejects us?
- We fix and re-file at no charge within the first 60 days.
- Do you touch payments?
- Never; Stripe does. A tool that needs money hands back a Stripe Checkout link and the payment happens there. No card number reaches Counter.
- How long does a build take?
- About five business days from the scoping call for a business with a working public API. Directory review is out of our hands: Claude and ChatGPT publish no service level, and Muse publishes no rules at all.
- What do you need from us?
- API documentation, a test account, and someone who can answer questions for an hour. For the Claude directory you also need a Claude Team or Enterprise organisation in your own name, because that is where the submission happens.
Start with the evidence
The scan is free and it scopes the build
Paste a domain, an OpenAPI URL, or an MCP URL. The report names what each of the three directories would reject today, with the fix beside each item.