AgenticVision Overview

AgenticVision provides persistent visual memory for agents through a portable .avis artifact and an MCP server.

What it does

• Captures images from files, base64 payloads, screenshots, and clipboard.
• Stores embeddings and metadata in .avis.
• Supports visual search, comparison, diff, OCR, and memory linking.
• Exposes tools to any MCP-compatible client through agentic-vision-mcp.

Why teams adopt AgenticVision

Teams adopt AgenticVision because it closes both legacy and current visual-intelligence gaps:

• Foundational problems already solved: no persistent visual evidence store, no queryable visual history, no structured visual comparison/diff workflow, no visual quality diagnostics, no vision-to-memory bridge, and no universal MCP visual runtime.
• New high-scale problems now solved: UI-state blindness during automation and non-text signal blindness in decision workflows.
• Practical outcome for teams: decisions are grounded in what the system actually saw, not only text logs, with portable evidence that can be queried later.

For a detailed before-and-after view, see Experience With vs Without.

Artifact

• Primary artifact: .avis
• Cross-sister server workflows can pair .avis with .amem and .acb

Start here

• Installation
• Quickstart
• Command Surface
• Runtime and Sync
• Integration Guide
• Experience With vs Without

Works with

• AgenticMemory — link visual captures to memory nodes with vision_link for cross-modal evidence trails.
• AgenticCodebase — pair .avis screenshots with .acb code graphs to connect UI regressions to code changes.