REST API · MCP · verifiable actions

Markdown

AI text watermark and metadata removal API

Gifi exposes the same inspect, clean, and rewrite pipeline used by the web app. Text cleaning returns named Unicode and punctuation changes. File cleaning reports each metadata action. Rewriting changes prose but never claims to certify a private detector.

Read the API reference Create an API key

Endpoints

One pipeline, four calls

MethodPathJob
POST/api/v1/inspectInspect text without spending a credit
POST/api/v1/cleanRemove named Unicode or file metadata
POST/api/v1/rewriteGenerate new wording for a statistical text signal
GET/api/v1/usageRead plan, credit, and rate-limit status

Inspect before you spend

Inspection is free. Clean and rewrite calls spend credits only on work that runs on Gifi’s machines, and failed jobs are refunded.

curl -X POST https://gifi.ai/api/v1/inspect \
  -H "Authorization: Bearer $GIFI_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"text":"Text to inspect"}'

Supported input

Text, PNG, JPEG, WebP, SVG, HTML, Markdown, PDF, DOCX, and ODT. Unsupported or incomplete cleaning paths are refused rather than returned as clean.

Hard limits

  • No API response certifies that text will pass or fail a vendor detector.
  • C2PA soft binding can re-link a cleaned file to a remote manifest.
  • Visible or pixel-embedded watermarks, audio, and video are outside the API scope.

API questions

Is this an AI detector API?
No. Inspect names verifiable characters and metadata. Rewrite changes wording. Gifi cannot see or certify a vendor’s private detector key.
Can the API remove visible image watermarks?
No. It removes named metadata containers, not logos or signals embedded in image pixels.
Can agents use the same tools?
Yes. The remote MCP server exposes inspect, clean, rewrite, and usage tools backed by the same account and credits.

Prefer tools to endpoints?

Connect the same account over MCP, or compare the credit plans before integrating.

Connect MCPSee pricing