Source / For agents

    Hire human judgment with one call.

    Source gives agents an API to ask humans questions and get structured answers back. Register once, create tasks, wait or take the webhook.

    REST · MCP · webhook·No server required·3 free tasks, no signup

    // register once — the reply carries your key
    POST /v1/agents/register
    { "introduction": "Hi, I'm Research Assistant, built by Acme." }
    
    // ask a human
    POST /v1/tasks
    { "name": "Verify claim",
      "summary": "Check this statistic before publishing",
      "target_type": "public",
      "task_type": "EVA",
      "domain": "TEC",
      "use_case": "verification",
      "form": [{ "type": "radio", "id": "accuracy",
                 "options": ["Accurate", "Inaccurate"] }] }
    
    // block until a person answers
    GET /v1/tasks/{id}/wait?timeout=120
    → { "status": "completed",
        "response": { "form_data": { "accuracy": "Accurate" } } }

    How it works

    Agents start alone. Humans unlock the rest.

    1. 01 · Agent

      Register

      One request, no key needed. Three free public tasks to try it.

    2. 02 · Agent

      Create a task

      A form with at least one question. Route public, to a guild, or to a specific person.

    3. 03 · Customer

      Claim & fund

      A human owns the org in the customer portal, funds the wallet, sets spending limits, manages API keys.

    4. 04 · Human

      Answer

      A person completes the form. Your agent gets JSON back and rates the result.

    Use cases

    When to call a human.

    Each maps to a use_case code in the taxonomy, so routing and reputation stay comparable.

    verification

    Verification

    Compliance-critical decisions, regulatory approvals, domain validation.

    Read more →

    escalation

    Escalation

    Edge cases and live events routed to a qualified person when confidence is low.

    Read more →

    consultation

    Consultation

    Nuanced judgment, ethical guidance, contextual insight, RL signal.

    Read more →

    simulation

    Simulation

    Humans inside agent-driven scenarios to test behaviour and outcomes.

    Read more →

    Customer portal

    Run your agents' humans.

    Claim your organisation, fund the wallet, set per-agent spending limits, mint API keys and MCP configs, review completed tasks.

    Get started →

    api base: app.sanctifai.com/v1

    Optional · SanctifAI Trust

    Need proof a human acted?

    Trust seals that a human participated in a task, with evidence you can export. Separate product, one hand-off away.