document-vision-mcp
Server Configuration
Describes the environment variables required to run the server.
| Name | Required | Description | Default |
|---|---|---|---|
| DOCUMENT_VISION_ROOT | No | The root folder containing PDFs. Single folder. Used if DOCUMENT_VISION_ROOTS is not set. | |
| DOCUMENT_VISION_ROOTS | No | JSON array string of folders containing PDFs. Overrides DOCUMENT_VISION_ROOT when set. Example: '["./ref-design", "../ortak-dokumanlar"]' |
Instructions
Guidance the server publishes about itself, which clients place ahead of the tool catalog so the model reads it before choosing anything.
This server publishes no instructions, or was last inspected before Glama recorded them.
Capabilities
Features and capabilities supported by this server
Protocol revision2025-11-25
| Capability | Details |
|---|---|
| tools | {
"listChanged": false
} |
| prompts | {
"listChanged": false
} |
| resources | {
"subscribe": false,
"listChanged": false
} |
| experimental | {} |
Tools
Functions exposed to the LLM to take actions
| Name | Description |
|---|---|
| list_documentsA | List PDF filenames and root mappings. Multiple roots use root1/, root2/ prefixes. Pass the returned filename unchanged to other tools. Single-root names have no prefix. |
| document_infoB | PDF page count, sizes and text-layer availability. Paginated, maximum 100 pages. |
| search_textA | Find literal PDF text (part numbers, net names); return page and normalized bbox. Search up to 200 pages per call. Expand hit bboxes before cropping surrounding circuitry. This is not semantic search or OCR. |
| page_textA | Extract text from one PDF page or normalized crop. Paginate with next_offset; no OCR. |
| page_renderA | Return full page as inline PNG plus metadata. Start at 120-180 dpi for an overview. For small labels, use page_crop. Never silently resizes. Limit: 12 MP / 6000 px / 8 MB PNG. |
| page_cropA | Re-render a PDF region at high resolution, returning inline PNG and metadata. bbox=[x0,y0,x1,y1], all 0..1, origin top-left of FULL visible page. Example: [0,0,0.5,0.5] is the upper-left quarter. dpi=36..600. Expand search hit bbox to include surrounding wires/components. Not an upscale of a thumbnail. |
Prompts
Interactive templates invoked by user choice
| Name | Description |
|---|---|
No prompts | |
Resources
Contextual data attached and managed by the client
| Name | Description |
|---|---|
No resources | |
TDQS
Scored across 6 tools
Each tool targets a distinct action: listing files, metadata, literal text search, page text extraction, full-page rendering, and region cropping. The only mild overlap is between page_render and page_crop, but descriptions clearly differentiate full-page overview versus high-res regional re-render.
Names mix conventions: verb_noun (list_documents, search_text), noun_noun (document_info, page_text), and noun_verb (page_render, page_crop). Shared document_/page_ prefixes aid readability, but the pattern is not uniform.
Six tools is well-scoped for a PDF reading/vision server. Each tool earns its place with no redundant operations, covering the natural read-only workflow without bloat.
The surface covers the full read lifecycle: discover files, inspect metadata, search text, extract text, render pages, and crop regions. No OCR is offered, but that is explicitly out of scope; coverage is otherwise solid for the stated purpose.