Smarter Code Intelligence
Kommit now understands your code at a structural level. Repositories are analyzed by an AI agent that maps architecture, dependencies, and features automatically — backed by a hardened pipeline that runs reliably at scale.
Highlights
- Code-aware embeddings split at function and class boundaries via AST parsing
- AI agent loop explores your codebase and builds an architecture graph
- Repository sync runs on dedicated infrastructure with live progress tracking
Assets, MCP & Canvas Polish
Upload images, videos, and PDFs directly to your canvas. Connect your AI coding tools via MCP with one-click API key generation. Plus a wave of canvas UX improvements.
Highlights
- File uploads with image annotation, video thumbnails, and version history
- MCP integration so Claude Code and Cursor can read your specs live
- Smarter canvas with auto-connecting nodes and contextual actions
PRD Generation & Export
Generate production-ready Product Requirement Documents directly from your canvas. AI synthesizes all your nodes into a structured, export-ready PRD.
Highlights
- AI-powered PRD generation from canvas nodes with streaming output
- Export to Markdown, PDF (branded), or structured JSON
- Version history with one-click regeneration
GitHub Connector
Connect your GitHub repositories directly to your canvas. Kommit ingests your codebase and uses it as context for smarter AI conversations and PRD generation.
Highlights
- GitHub App integration with one-click repo connection
- Smart file filtering and code-aware chunking for context
- Auto-detection of tech stacks and database schemas
Knowledge & RAG
Your project canvas now has a memory. Every conversation and data point is embedded and retrieved contextually, so the AI always knows what you've discussed before.
Highlights
- Vector embeddings for semantic search across your project
- Contextual retrieval with footnote-style citations
- Lazy backfill ensures existing projects get embeddings automatically
AI Conversation Engine
Every node on your canvas now has its own AI conversation. The AI guides you through filling out structured requirements while understanding context from connected nodes.
Highlights
- Per-node AI chat with real-time streaming responses
- Inline structured data extraction as you talk
- Cross-node context injection for holistic understanding
Foundation
The initial release of Kommit. A visual canvas for organizing product specs as connected nodes, with authentication, multi-tenancy, and auto-saving.
Highlights
- Visual canvas with 6 node types for structuring product specs
- Authentication with GitHub OAuth and email
- Secure workspace isolation with per-org data boundaries
