First-class PDF tool with native Anthropic and Google support. Extract text, analyze documents, and route through your preferred provider.
Native PDF analysis
OpenClaw now has a first-class PDF tool so your assistant can read, analyze, and extract information from PDF documents.
Provider support
- Anthropic — Native PDF input (Claude 3.5+)
- Google — Native PDF via Gemini
- Fallback — Non-native models get text extraction; content is passed as plain text
Configuration
You can tune defaults in your config:
{
"agents": {
"defaults": {
"pdfModel": "anthropic/claude-sonnet-4-5",
"pdfMaxBytesMb": 20,
"pdfMaxPages": 50
}
}
}
pdfMaxBytesMb— Max file size (default 20 MB)pdfMaxPages— Max pages to process (default 50)
Use cases
- Research — Summarize whitepapers, extract key findings
- Contracts — Review terms, extract dates and parties
- Reports — Analyze financial or project reports
- Manuals — Answer questions from documentation
Diff output as PDF
The Diffs tool now supports PDF output for generated diff artifacts — useful when messaging channels compress images. Configure fileQuality, fileScale, and fileMaxWidth for rendering.