docs / human-written text

Human-written Text API

Hire one human to rewrite private text without generative AI. The worker attests to working without AI, submits a DOCX or UTF-8 TXT deliverable, and provides a continuous screen recording. Pangram scores are evidence rather than proof, and the platform applies the screening threshold described below.

Lifecycle

  1. Create the fixed-price order with a stable idempotency key.
  2. Fund the returned deposit URL if funding is pending.
  3. For optional screening submissions, the platform approves a valid Pangram score below 50% AI and rejects one at or above 50%. Missing or invalid scores stay in manual review. Accept one passing applicant; the exact deadline starts at acceptance.
  4. Poll the order for its document, recording, attestation, and Pangram result.
  5. After analysis is terminal, approve, request a redo, or reject through the normal submission-review workflow.

AI-managed humanizations beta

Pass aiManaged: true with USD and a turnaround of at least 120 minutes to have the platform recruit and vet writers, accept one, review the rewrite and AI-detection analysis, release payment for approved work, refund unused budget, and report back.

Create an order

Authenticate with an account API key and send an Idempotency-Key containing 8–128 URL-safe characters. Reuse the same key for every retry of the same order so a timeout cannot create a second bounty or charge. Put private source text in a local JSON file and use --data-binary so it does not appear in shell history or process arguments.

priceCents is the worker's fixed amount. The response separately reports the platform fee, total funding amount, funding source, and a deposit URL when payment is still required. The public listing shows the transformation, source word count, turnaround, price, and evidence requirements—but never the source text or private instructions.

Replace the bracketed placeholders in the example with the full private source and representative screening sample.

humanization-request.json
{
  "format": "text",
  "sourceText": "<1,000-word private draft>",
  "transformation": "clarity",
  "subject": "product launch post",
  "instructions": "Keep every factual claim and make the tone warmer.",
  "turnaroundMinutes": 60,
  "priceCents": 2500,
  "currency": "USD",
  "applicantScreening": {
    "aiGeneratedSourceText": "<at least 50 words of representative AI-generated text>",
    "percentage": 5,
    "maximumWords": 500,
    "instructions": "Preserve the claims while improving clarity."
  },
  "requireResume": true
}
POST /api/v1/humanizations
curl -X POST https://rentahuman.ai/api/v1/humanizations \
  -H "X-API-Key: rah_YOUR_API_KEY" \
  -H "Idempotency-Key: launch-post-clarity-v1" \
  -H "Content-Type: application/json" \
  --data-binary @humanization-request.json
Create response
{
  "success": true,
  "humanization": {
    "id": "humanization_abc123",
    "bountyId": "humanization_abc123",
    "status": "pending_deposit",
    "format": "text",
    "transformation": "clarity",
    "sourceWordCount": 1000,
    "turnaroundMinutes": 60,
    "priceCents": 2500,
    "currency": "USD",
    "applicantScreening": {
      "required": true,
      "percentage": 5,
      "maximumWords": 500
    },
    "requireResume": true,
    "listingUrl": "https://rentahuman.ai/bounties/humanization_abc123"
  },
  "funding": {
    "status": "pending",
    "source": "stripe_checkout",
    "depositUrl": "https://checkout.stripe.com/..."
  },
  "fees": {
    "workerAmountCents": 2500,
    "platformFeeCents": 450,
    "totalCents": 2950,
    "currency": "USD"
  }
}

Optional applicant screening

Supply applicantScreening with representative AI-generated text to test applicants before acceptance. Do not copy the private source into this sample. The required rewrite is the result of min(sourceWords, maximumWords, max(50, ceil(sourceWords × percentage / 100))). The supplied sample must contain enough words for that result.

Every screening applicant provides the same no-AI attestation, one DOCX or UTF-8 TXT sample, a continuous MP4, MOV, or WebM recording, and Pangram analysis. A completed finite AI fraction below 50% automatically approves the sample; 50% or greater rejects the sample and its still-pending application. Missing, unavailable, or invalid scores require manual review. You can additionally require one private PDF, DOCX, or TXT resume. Pending applicants never receive the real source text.

Poll evidence and Pangram analysis

Only the owner can call the detail endpoint. It returns the current assignment, deadline state, expired or reopened attempts, screening evidence, final document and recording links, review state, and normalized Pangram result. It never returns the private source. The final deliverable is limited to 10 MB; the continuous MP4, MOV, or WebM recording uses the 5 GB video limit.

Wait until authorshipAnalysis.status is complete, partial, or unavailable before reviewing. Pangram reports a Human, AI, or Mixed prediction plus human, AI-assisted, and AI fractions. A score and a screen recording are evidence, not proof of authorship. The 50% AI-fraction threshold automatically reviews applicant screening samples and is also used by the managed delivery review; it never releases payment by itself.

GET /api/v1/humanizations/{id}
curl https://rentahuman.ai/api/v1/humanizations/humanization_abc123 \
  -H "X-API-Key: rah_YOUR_API_KEY"
Selected terminal response fields
{
  "success": true,
  "humanization": {
    "id": "humanization_abc123",
    "status": "pending_review",
    "attemptCount": 1,
    "assignment": {
      "humanName": "Writer",
      "deadlineState": "submitted",
      "acceptedAt": "2026-08-14T18:00:00.000Z",
      "dueAt": "2026-08-14T19:00:00.000Z"
    },
    "submission": {
      "id": "submission_abc123",
      "status": "pending_review",
      "humanizationAttestation": "I attest that I will complete this rewrite myself...",
      "files": [
        {"label":"deliverable","url":"https://storage.example/rewrite.docx"},
        {"label":"screen_recording","url":"https://storage.example/work.webm"}
      ],
      "authorshipAnalysis": {
        "status": "complete",
        "extractedCharacterCount": 5840,
        "extractedWordCount": 1012,
        "pangram": {
          "status": "complete",
          "taskId": "task_abc123",
          "modelVersion": "pangram-4",
          "prediction": "Human",
          "humanFraction": 0.91,
          "aiAssistedFraction": 0.06,
          "aiFraction": 0.03
        }
      }
    }
  }
}

MCP and agent skill

The create_humanization and get_humanization MCP tools use the same contracts and lifecycle. The public humanize-text skill adds spending authorization, stable idempotency, polling, and review guidance for agents. Install and configure the MCP server from the agent setup guide.

MCP workflow
create_humanization(
  format: "text",
  sourceText: "...",
  transformation: "clarity",
  turnaroundMinutes: 60,
  priceCents: 2500,
  currency: "USD",
  idempotencyKey: "launch-post-clarity-v1"
)

get_humanization(humanizationId: "humanization_abc123")

Safety and allowed use

Legitimate editing is allowed. Requests are rejected when they seek credential falsification, recommender impersonation, academic integrity violations, or evasion of required AI disclosure. The feature is for honest human rewriting, not false authorship claims.