Call Clearance
Pre-call compliance for AI voice and SMS agents

Every outbound call cleared
before it's placed.

One question, asked by your dialer or by the agent itself, before a call is placed. It checks consent on file, the national and state Do Not Call lists, known litigators, calling hours and opt-outs, and returns allow or block in under 200 ms. Every decision is logged. If a lawyer asks, you have the record.

Request access See the integration For agencies running outbound on Retell, Vapi, Bland or ElevenLabs.
Retell Vapi Bland ElevenLabs Agents Twilio SMS Any agent that uses tools
One check, before one call
+1 410 555 0123for Harbor Dentalvoice
  1. Opt-out listclear
  2. Litigator listclear
  3. National Do Not Callclear
  4. State Do Not Callclear
  5. Consent on fileweb form, Sep 2
  6. Calling hours10:00 local
Allow. Dial. Logged as evidence record ev_01J9K3ZQ7V, 41 ms.

A voice agent dials faster than anyone checks the list.

TCPA damages are $500 to $1,500 per call, and most TCPA suits are class actions. Retell, Vapi and ElevenLabs say in their own documentation that consent and Do Not Call compliance are your problem. Call Clearance is the part of your stack that handles it.

Consent of record

Post each opt-in to one ledger with its source, timestamp and evidence: web forms, text replies, recorded verbal consent, written consent. The pre-call check reads this ledger first.

Do Not Call and litigators

The national registry under each client's own FTC account, the state lists, and a known-litigator list, refreshed on schedule. Each client is registered separately, which is what the FTC requires.

Evidence you can hand over

Every decision goes into a hash-chained log that shows any edit. Export a per-number report with the consent record and every check that ran on it, as a PDF or a data file. Your records stay yours.

What runs before each dial

Six checks, in this order. Every result, pass, hit or skipped, is written to the log with the reason. The whole pass is one round trip and one log entry.

1Opt-outAnyone who said stop, across every client on your account.
2Litigator listNumbers tied to serial TCPA plaintiffs and demand letters.
3National DNCUnder the client's own FTC subscription. Reports as skipped until that's linked.
4State DNCState registries, keyed off the number's area code.
5Consent on fileAn active consent record for this number, this client, this channel.
6Calling hoursLocal time for the number against the allowed window.
1

Add your clients

Each business you call for gets its own account. We handle the FTC registration and state-list setup for each one, so every scrub runs under the right account.

2

Record consent as it happens

Post opt-ins to the ledger from your forms, your SMS replies, your agent or your CRM.

3

Check before every call

Your dialer or the agent itself sends the number and client. It gets back allow or block with a reason.

4

Propagate every opt-out

When a contact says stop, the opt-out is posted and the number is blocked for every client on your account from that moment.

The whole integration

Two ways in, one gate. A dialer asks with a single web request. An agent asks as a tool, with descriptions that tell the model to check before every dial.

Your dialer or agent asks Call Clearance before each call and gets allow or block; every answer is logged. Your dialer one web request before dialing Your voice agent asks as a tool, on Retell or Vapi number + client CALL CLEARANCE 1 Opt-out list 2 Litigator list 3 National Do Not Call 4 State Do Not Call 5 Consent on file 6 Calling hours under 200 ms, one log entry allow place the call block, error or timeout do not dial, reason recorded hash-chained audit log, exportable
  1. Your dialing code sends the number and the client it's calling for, with your key.
  2. The answer comes back as allow or block, with the reason and an evidence ID.
  3. Dial only on allow. Treat block, an error or a timeout as do not dial.
  4. Post consent and opt-outs the same way, as they happen. Ask for any number's evidence file when someone wants proof.

Works from any language, any workflow tool, or the pre-call hook on your platform. Step by step on the setup page; the technical reference is at api.callclearance.com/docs.

  1. Add Call Clearance to the agent as a tool server. One address, one key. That's the whole setup.
  2. The agent gets five tools: check a number, record consent, record an opt-out, fetch evidence, list clients.
  3. The tool descriptions tell the model to check before every dial, so it does. It can ask, record and fetch. It cannot change the rules; those stay on your account, set by a person.
  4. Same key and same one cent per check as the dialer path.

Any MCP client works, including Retell and Vapi agents, Claude, and your own agent stack.

  1. In the assistant's tools, add an MCP server. Paste the Call Clearance address and your key as the authorization header.
  2. In the assistant's prompt, add one rule: check the number before any outbound call, dial only on allow, and stop on block or error.
  3. Run a test call to a blocked test number and confirm the assistant declines to dial.

If your dialer places the call before the assistant runs, use the dialer path instead.

  1. In the agent's tools, add an MCP server with the Call Clearance address and your key. Pick the check, consent and opt-out tools.
  2. Add the same rule to the agent prompt: check before every dial, dial only on allow, treat block or error as do not dial. Retell doesn't retry a failed tool call, so the prompt rule is what keeps a failed check from turning into a dial.
  3. Test with a blocked test number before going live.

For campaign dialing from your own code through Retell's API, use the dialer path.

What the evidence looks like

When someone asks "was this number OK to call," you send them one file. It has the consent record, every check that ran, the decision, and a chain hash that shows nothing was edited afterwards.

  • Per-number reportEvery check and decision for that number under that client, in order, with timestamps.
  • Consent recordChannel, source, when it was captured, and a hash of the evidence you attached.
  • Hash chainEach entry carries the hash of the one before it. Change any entry and the chain breaks.
  • PDF or dataA PDF for the person asking, a data file for your systems. Same content.
Consent, checks and opt-outs are written as linked entries; each carries the hash of the one before it, and the whole chain exports as one report. Consent recorded, Sep 2 web form, voice, evidence attached and hashed 4b77…0c9d each entry carries the hash of the one above Check, Sep 11 six checks ran, all clear allow 9f3c…e21a Opt-out recorded, Sep 14 said stop on a call; blocked for every client from now on c41d…77b0 Check, Sep 15 stopped at the opt-out list, nothing dialed block 7a19…d3e4 One report per number, as PDF or data

Pricing

A monthly fee plus a per-check rate. No seats, no annual contract, no per-client fees.

Agency

$99 / month
plus $0.01 per check
  • Unlimited clients
  • Consent ledger and evidence export
  • National, state and litigator lists
  • Dialer path and agent path, same key
  • FTC registration handled for each client

Volume

Talk to us
for more than 500,000 checks a month
  • Lower per-check rate
  • Review of your consent flows
  • State telemarketing registrations
  • Custom retention and export
$1,099per month, every call checked

Questions people ask first

Short answers. Anything else, email.

Does this make me compliant?

No tool does. It runs the checks a reasonable outbound program is expected to run, before every call, and keeps proof that it did. What counts as consent for your campaigns is still a question for your counsel. We give you the record; they give you the advice.

Whose FTC account is the national list scrubbed under?

Each client's own. The FTC requires the seller to hold the subscription, so we set one up per client and run their scrubs under it. Until a client's subscription is linked, the national check reports as skipped in the answer and in the log, not as clear. We don't hide it.

What happens if the check is slow or down?

Your code or your agent should treat an error or a timeout as do not dial, and the MCP tool descriptions say exactly that. A missed call is cheap. A call you can't defend isn't.

Can the agent change the rules?

No. The MCP tools let an agent ask, record and fetch. Consent policy, calling windows and list settings live on your account and are changed by a person, not by a prompt.

Who owns the records?

You do. Export any number, any client, or everything, at any time, as PDF or data. If you leave, you take the ledger with you.

How fast can I start?

Onboarding is by hand right now. Email with your platform and monthly call volume. You get a key the same week, and the check works on real numbers from the first call.

Request access

We're onboarding a small number of agencies by hand. Email us with the platform you run on and how many outbound calls you make a month.

[email protected]