Build with clear outcomes

Help and API guide

Define a small set of questions, submit communication, and review typed answers before your application takes action.

Write an assessment

Start with one H1 title. Add one H2 for each question, then use H3 headings for that question’s fields. Put each field value after a colon or on the next line.

# Follow-up

## Does this message need a reply?

### Fact ID: reply_needed
### Type: boolean

## What should happen next?

### Fact ID: next_action
### Type: choice
### Choices: reply, schedule, close

Each question needs a stable Fact ID and a supported type: boolean, choice, or bounded score. Add a comma-separated Choices list for choice questions and the supported minimum and maximum for a score. For a bounded score, add ### Min: 1 and ### Max: 5. You can also set ### Max input tokens: 2048 and ### Max output tokens: 96. Keep questions focused. A definition can have up to eight questions.

Run an assessment

Send a JSON request to POST /v1/assessments/run. Keep your API key in a server-side secret store. Never place it in browser code or a public repository.

curl https://app.closingthreads.com/v1/assessments/run \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -H "Idempotency-Key: YOUR_UNIQUE_REQUEST_ID" \
  -d '{
    "communication": "Thanks, I will send the estimate Thursday.",
    "markdown": "# Follow-up\n\n## Does this need a reply?\n\n### Fact ID: reply_needed\n### Type: boolean"
  }'

Use a new unique idempotency key for each request. If a request times out, reuse its original idempotency key with identical input. A repeat returns the recorded status, not a saved copy of the result. Check your account activity; an unknown outcome needs operator review before its held credit can be released. Using a new key can incur another charge. The service returns typed results, a request ID, the model and pricing version, measured token use, and the charge. Results don’t include a calibrated confidence score. Review each answer before acting on it.

Privacy and sensitive content

Private mode is the default. Closing Threads does not store request content or results after processing, but the inference provider receives the request. Standard mode stores content encrypted for up to 30 days only with your explicit consent. Review provider privacy terms before you submit sensitive data.

Billing

When purchases are available, you can buy US$10, US$25, or US$100 in prepaid credits. Applicable tax is calculated at checkout and is separate from your credit balance. Spending stops when your available credit or monthly limits are reached.

Get help

Email colin@nixc.us with your request ID if an outcome stays unknown. Don’t send API keys, recovery codes, or sensitive request content.

Passkeys and recovery

Save your recovery codes and account ID in a safe place. Each recovery code works once. Generate a new set after using or losing codes; doing so invalidates the previous set. Add more than one passkey so you have another sign-in option.