Overview
This release delivers the most-requested feature, CLI and API access to your Edit Mind library (closes #109), alongside an MCP server for AI assistant integration, access token management, and a batch of reliability fixes.
Features
REST API (/api/v0)
Edit Mind now exposes a public REST API from the background-jobs service so you can query your library programmatically:
| Endpoint | Description |
|---|---|
GET /api/v0/videos
| List videos with metadata and thumbnails |
GET /api/v0/videos/:id
| Get a single video with all its scenes |
POST /api/v0/search
| Semantic search across video scenes |
GET /api/v0/collections
| List AI-generated collections |
GET /api/v0/collections/:id
| Get a collection with its video items |
GET /api/v0/suggestions
| Search suggestions and popular queries |
GET /api/v0/media
| Serve media files (images, thumbnails) |
All routes are protected by Bearer token authentication with scope-based authorization.
MCP Server
A standalone MCP server (edit-mind-mcp) wraps the REST API and exposes your video library as tools for AI assistants like Claude and Cursor:
search_video_scenes— semantic search with location and aspect ratio filterslist_videos/get_video— browse and inspect your librarylist_collections/get_collection— explore AI-generated collectionsget_suggestions— discover available contentview_media— display thumbnail images inline in the AI chat
The MCP binary is bundled into a single file and attached to every release. To add it to Claude Code, run the install script:
curl -fsSL https://raw.githubusercontent.com/IliasHad/edit-mind/main/scripts/install-mcp.sh | bashAccess Token Management
Create and manage personal access tokens from the Settings page. Each token supports:
- Custom name with optional expiry date
- Fine-grained scope selection
- Last-used tracking (timestamp, IP, user-agent)
Install Script
A new scripts/install-mcp.sh makes it easy to set up Edit Mind and its MCP server on your machine or server.
Bug Fixes
- Video transcoding codec errors now surface in the UI with a clear failure reason instead of silently failing (#130)
- Video chat results no longer fail to render due to a race condition when loading video metadata (#131)
- Job progress tracking added for embedding and analysis stages for better visibility into processing status
- Worker reliability improved with better lock management, stall detection, and sequential embedding processing
- Model selection fixed to correctly prioritize backends
- Path validator issues resolved
Full Changelog: v0.21.0...v0.22.0