Skip to main content
RigorDesk logoRigorDesk
Start free

Quickstart

Start with a draft or bring work you already have. You can also configure a panel first if you know which models and format you want.

In the workspace

  1. Create an account or sign in to open your workspace.
  2. Choose Create a draft to write with an AI model, or Review existing work to paste a proposal, plan, or AI answer. Add the original question and relevant context.
  3. Open the draft or response you want reviewed and choose Review. Choose the review goal, format, and available models. Check the settings before starting.
  4. Read the conclusion alongside the reviewers' responses and unresolved objections. Use the findings to revise your draft or identify evidence to check.

For example, bring a proposal to switch a service to annual billing. Include the monthly price, annual price, cancellation terms, and expected period of use. Ask whether the savings justify the commitment.

Hobby includes two Refinement reviews each month. These reviews use free-tier models and run up to two rounds each. Compare plans for other formats and higher limits.

To start with the reviewers instead of a draft, choose Set up a review on the workspace home screen. This opens the panel editor and prompt workspace.

With the API

Use the API when your own application should submit a question and retrieve the review. API access requires a paid subscription. See pricing for tier limits.

1. Get an API key

  1. Sign in at /auth.
  2. Go to Settings → API Keys.
  3. Enter a name such as prod-backend, then click Create key.
  4. Copy the token. It starts with dbt_ and is shown only once.

Anyone with your key can run debates and spend your budget. Don't commit keys to git. Rotate by revoking the old key and creating a new one.

2. Create an assembly

An assembly is a saved panel configuration. Create it once and reuse it across reviews. The API calls a review a debate.

curl -X POST https://rigordesk.com/api/v1/assemblies \
 -H "Authorization: Bearer dbt_YOUR_KEY" \
 -H "Content-Type: application/json" \
 -d '{
 "name": "My first panel",
 "models": [
 { "id": "openai/gpt-5.5", "role": "INITIATOR" },
 { "id": "anthropic/claude-opus-4.8", "role": "REVIEWER" },
 { "id": "x-ai/grok-4.3", "role": "SYNTHESIZER" }
 ],
 "maxRounds": 3,
 "agreementLimit": 75
 }'

The API returns a JSON object with an id. Save it. That value is your assemblyId for every future debate.

3. Run a review

curl -X POST https://rigordesk.com/api/v1/debates \
 -H "Authorization: Bearer dbt_YOUR_KEY" \
 -H "Content-Type: application/json" \
 -H "Idempotency-Key: $(uuidgen)" \
 -d '{
 "prompt": "Should new 2026 projects start with low-code + AI agents, or hand-written code?",
 "assemblyId": "PUT_THE_ASSEMBLY_ID_HERE"
 }'

The API returns 202 Accepted with {"id": "...", "status": "queued"}. Save that id. It is your sessionId.

4. Poll for the result

curl https://rigordesk.com/api/v1/debates/PUT_THE_SESSION_ID_HERE/status \
 -H "Authorization: Bearer dbt_YOUR_KEY"

Keep polling every few seconds. When isComplete is true, fetch the full transcript:

curl https://rigordesk.com/api/v1/debates/PUT_THE_SESSION_ID_HERE \
 -H "Authorization: Bearer dbt_YOUR_KEY"

The full response includes every round, every model's turn, the per-turn cost, and the chairman's finalAnswer at the top.