Optical Context MCP
Feed long OCR-heavy PDFs to vision-capable agents by compressing pages into dense packed images that fit context windows.
Overview
Optical Context MCP is a MCP server for the Build phase that compresses OCR-heavy PDFs into dense packed images so agents can work with long visual documents.
What is this MCP server?
- Optical Context MCP v0.1.4 distributed via PyPI package optical-context-mcp
- Transforms OCR-heavy PDFs into dense packed images for long visual documents
- stdio MCP transport for local agent hosts
- Bridges scanned manuals, filings, and spec PDFs that blow text-only token budgets
- Open-source repo on GitHub for ChrBoebel/optical-context-mcp
- Server version 0.1.4 on PyPI identifier optical-context-mcp
- Repository: github.com/ChrBoebel/optical-context-mcp
Community signal: 1 GitHub stars.
What problem does it solve?
Huge scanned PDFs either truncate in the context window or cost a fortune when you OCR them page by page in chat.
Who is it for?
Builders creating document-QA or review agents over scanned specs, contracts, or manuals in Claude Code or Cursor.
Skip if: Simple text-native PDFs that extract cleanly with a plain text parser, or workflows with no vision-capable model.
What do I get? / Deliverables
After install, your agent can request packed optical views of long PDFs and analyze full documents within vision-model limits.
- Packed image representations of OCR-heavy PDFs via MCP tools
- Agent sessions that retain full-document visual context within limits
- Less manual chunking when building document analysis features
Recommended MCP Servers
Journey fit
Optical context tooling lands in Build when you are wiring document-heavy workflows into your agent stack. Packing PDFs for multimodal models is agent-tooling: it changes how the agent ingests evidence, not how you ship marketing.
How it compares
MCP document packing for vision models, not a generic PDF-to-markdown converter skill.
Common Questions / FAQ
Who is Optical Context MCP for?
Solo developers building agent workflows over long scanned or OCR-heavy PDFs who need vision-friendly context packing.
When should I use Optical Context MCP?
Use it during build when integrating multimodal agents that must read entire visual documents without exceeding context limits.
How do I add Optical Context MCP to my agent?
Install optical-context-mcp from PyPI, configure a stdio MCP server entry in your host, and expose the packing tools to your agent session.