API documentation

Version 1 · Base URL https://probator.ai/v1

Run Probator's checks from your own software: AI detection (with our own detection model, a rewrite test and an expert reading), grammar and style suggestions, plagiarism and hidden-mark forensics. The API uses the same credits as your plan.

Overview

• API access is included in the Pro and Team plans.
• Send text as JSON, or upload a file (PDF, DOCX, TXT, MD, HTML, up to 20 MB) as multipart/form-data.
• Every response is JSON. Successful checks return the full report; errors return {"detail": "...", "code": "..."}.
• CORS is enabled, but never put an API key in code that runs in a user's browser. Call the API from your server.

Authentication

Create a key in the editor under Account → API keys. The full key is shown once. Send it in the Authorization header:

Authorization: Bearer pb_live_xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx

Keys draw credits from the account that created them. Revoke a key immediately if it leaks.

Requests

JSON body fields:

With multipart/form-data, send the same fields as form fields (booleans as "true") plus a file part. Check flags can be sent directly as ai, grammar, plagiarism, deep.

POST/v1/analyze

Runs any combination of checks in one request. AI detection is on by default.

curl https://probator.ai/v1/analyze \
  -H "Authorization: Bearer $PROBATOR_KEY" \
  -H "Content-Type: application/json" \
  -d '{"text": "Paste the text to check here…", "checks": {"ai": true, "grammar": true}}'

# Upload a file instead
curl https://probator.ai/v1/analyze \
  -H "Authorization: Bearer $PROBATOR_KEY" \
  -F file=@essay.pdf -F ai=true -F plagiarism=true

import os, requests

r = requests.post(
    "https://probator.ai/v1/analyze",
    headers={"Authorization": f"Bearer {os.environ['PROBATOR_KEY']}"},
    json={"text": open("essay.txt").read(), "checks": {"ai": True, "grammar": True}},
    timeout=120,
)
r.raise_for_status()
report = r.json()
print(report["verdict"]["label"], report["verdict"]["p_ai"])
print("credits left:", r.headers.get("x-credits-remaining"))

const res = await fetch("https://probator.ai/v1/analyze", {
  method: "POST",
  headers: { Authorization: `Bearer ${process.env.PROBATOR_KEY}`, "Content-Type": "application/json" },
  body: JSON.stringify({ text, checks: { ai: true, plagiarism: true } }),
});
if (!res.ok) throw new Error((await res.json()).detail);
const report = await res.json();
console.log(report.verdict.p_ai, report.plagiarism?.similarity);

Shortcut endpoints

Each runs a single check and accepts the same body (without checks):

GET/v1/usage

Returns your plan and credit balance:

{
  "plan": "pro",
  "credits": { "allowance": 500000, "used": 18250, "bonus": 0, "remaining": 481750,
               "period_start": 1791244800000, "period_end": 1793836800000 }
}

Response

The report contains the sections for the checks you ran. Character offsets (start, end) refer to the text field of the response, which is your text with invisible characters removed and look-alike letters normalised.

AI-detection results are probabilities, not proof. If you show them to your users, follow our AI Policy.

Credits and headers

AI detection costs 1 credit per word, grammar 1 per 3 words, plagiarism 2 per word, and a deep scan adds 1,500. Failed checks are not charged, and an identical request repeated within 24 hours is served from cache for free (not for plagiarism). Each response includes:

• x-credits-used: credits charged for this request
• x-credits-remaining: balance after the request
• x-cache: hit or miss

MCP server for AI agents

Probator is also a remote Model Context Protocol server, so assistants such as Claude, ChatGPT, Cursor or VS Code agents can run checks directly. It uses your API key and your plan's credits, exactly like the REST API.

Claude Code

claude mcp add --transport http probator https://probator.ai/mcp \
  --header "Authorization: Bearer $PROBATOR_KEY"

Cursor, VS Code and other clients (MCP configuration file)

{
  "mcpServers": {
    "probator": {
      "type": "http",
      "url": "https://probator.ai/mcp",
      "headers": { "Authorization": "Bearer pb_live_YOUR_KEY" }
    }
  }
}

Clients that only run local servers can connect through mcp-remote:

npx mcp-remote https://probator.ai/mcp --header "Authorization: Bearer pb_live_YOUR_KEY"

Keep the key out of shared configuration files. Results returned to the agent include the same "probabilities, not proof" notice as the editor.

Errors

Limits

• 60 requests per minute per key.
• Up to 60,000 words per check on paid plans.
• Files up to 20 MB.
• Requests can take 5–60 seconds depending on length and checks; use a timeout of at least 120 seconds.

Terms

Use of the API is subject to the Terms of Service, the AI Policy and, if you process other people's personal data, the Data Processing Agreement. Questions: hello@probator.ai.