Vision MCP Tool Specification

Python API: API_SPECIFICATION.md
Module SPEC: SPEC.md
Agent guidance: AGENTS.md

Version: v1.3.0 | Status: Not MCP-exposed | Last Updated: June 2026

Current MCP Surface

codomyrmex.vision does not currently expose production MCP tools. Its public surface is Python-first and uses local VLM/PDF backends directly.

Surface	Purpose
`VLMClient`	Analyze images through local Ollama VLM models
`PDFExtractor`	Extract text from PDF documents
`AnnotationExtractor`	Extract structured visual annotations
`VLMConfig`, `VLMResponse`, `Annotation`, `BoundingBox`, `PageContent`	Data models

If MCP tools are added, they should clearly document file-path handling, backend availability checks, and error shapes for missing local models.