Vreko is an intelligence platform for AI-native development. It runs as a local service, observes changes your AI tools make to your codebase, and builds a model of your code’s specific failure patterns - which files are fragile, which changes cascade, which AI tools tend to make which mistakes.
That intelligence is delivered back to your AI agent automatically, through the context files your agent already reads - structured as the workspace.json open standard. No tool calls. No agent configuration. Your agent sees what Vreko sees.
Vreko keeps local restore points as a safety net while it observes - but recovery is a side effect. Intelligence is the product.
Your code never leaves your machine. Only metadata reaches our servers, and only for Pro features you’ve opted into.
Quick Start
Install in 2 minutes. Your agent sees Vreko's intelligence on the next prompt.
Architecture
How Vreko observes, learns, and delivers - technically.
Integrations
Rules-file delivery, pre-tool-use hooks, MCP, CLI. Pick the most ambient option your setup supports.
CLI Reference
vr commands for agent harnesses and power users.
The Journey
| Day | What Happens |
|---|---|
| Day 1 | Per-tool AI attribution from the first session |
| Day 30 | Learns YOUR codebase’s fragile files and co-change patterns |
| Day 90 | Your agent sees Vreko’s intelligence on every prompt - automatically |
Your code never leaves your machine. Privacy details →
Built on the workspace.json open standard · By Marcelle Labs