RTL document translation for AI agents

# Submit a document for translation
curl -X POST https://docsrtl.com/v1/translate \
  -H "X-API-Key: drtl_your_key" \
  -F "[email protected]" \
  -F "target_lang=Arabic"

# → {"job_id":"abc123","status":"pending"}

# Poll until done
curl https://docsrtl.com/v1/jobs/abc123 \
  -H "X-API-Key: drtl_your_key"

# → {"job_id":"abc123","status":"completed"}

# Download the translated file
curl https://docsrtl.com/v1/jobs/abc123/download \
  -H "X-API-Key: drtl_your_key" \
  -o translated.docx

Base URL: https://docsrtl.com/v1 - Pro plan required for API access.

Drop our skill.md into any Claude Code project and the agent will know how to translate documents through DocsRTL - no setup, no integration code. The file is self-contained: API endpoints, auth, supported languages, error codes, and a reference Python implementation.

Install it into a project:

# Save into your project's skills directory
mkdir -p .claude/skills/docsrtl
curl -o .claude/skills/docsrtl/SKILL.md https://docsrtl.com/skill.md

# Set your API key
export DOCSRTL_API_KEY=drtl_your_key

Claude Code auto-loads skills from .claude/skills/. Once installed, ask the agent to translate any document - it will pick the right endpoint, poll, and save the result.

Use the trimmed OpenAI-compatible function spec:

# Fetch the function spec
curl https://docsrtl.com/v1/openapi-functions.json

Or paste this tool definition directly (matches /v1/openapi-functions.json):

{
  "type": "function",
  "function": {
    "name": "translate_document",
    "description": "Translate a Word document between any pair of supported languages while preserving all formatting. Works for LTR-to-RTL, RTL-to-LTR, LTR-to-LTR, and RTL-to-RTL. RTL targets (Hebrew, Arabic, Persian, Urdu) automatically receive right-to-left layout. Returns a job_id - poll get_job_status until completed, then call download_translation.",
    "parameters": {
      "type": "object",
      "properties": {
        "file_bytes_b64": {
          "type": "string",
          "description": "Base64-encoded document bytes (.docx, .doc, .odt, .rtf, max 50 MB)."
        },
        "filename": {
          "type": "string",
          "description": "Original filename including extension (e.g. report.docx)."
        },
        "target_language": {
          "type": "string",
          "description": "Target language name. Call list_languages() to get valid values - 40+ languages supported, not just RTL."
        },
        "translation_mode": {
          "type": "string",
          "enum": ["professional", "full", "literal"],
          "default": "professional",
          "description": "Translation quality mode: 'professional' (preserves technical terms), 'full' (immersive), 'literal' (word-for-word)."
        }
      },
      "required": ["file_bytes_b64", "filename", "target_language"]
    }
  }
}

For URL-based input use POST /v1/translate-url with body { url, target_lang }. GPT Actions spec available at /v1/openapi-gpt-actions.json.

A Python SDK with LangChain and LlamaIndex tool wrappers is in development. For now, call the REST API directly using httpx, requests, or any HTTP client.

Supported target languages