Vinv
VinvAI runs and observes your services to find bugs and perf issues
Last verified:
What is Vinv?
Vinv is a runtime debugging and verification tool that connects runtime traces to source code, feeds evidence to AI coding agents for fix generation, and automatically verifies solutions through acceptance tests. It runs 100% locally with zero code changes and integrates with Claude Code, Cursor, and any MCP-compatible agent.
Vinv pricing
Pricing model: Freemium
Free and open source (Apache 2.0)
Vinv pros
- 100% local execution with no cloud dependency, API keys, or telemetry
- Runtime evidence demonstrably improves AI performance (beats frontier models without context)
- Automatic fix verification and rollback via acceptance tests—fixes only land if they pass
- One-click VS Code install; also works as CLI or MCP client with any agent
- Smart Thompson sampling optimizes runtime context passed to agent for efficiency
- Free and open source (Apache 2.0)
Vinv cons
- Python-first; TypeScript and Go support listed as 'next'
- Requires ~500MB one-time embedding model download plus Rust and uv installation
- Requires acceptance tests to exist beforehand for verification to function
- Limited to services and APIs (not general-purpose code)
Frequently asked questions about Vinv
What editors does Vinv support?
VS Code, Cursor, Windsurf, VSCodium, Trae, VSCode Insiders, or any MCP-compatible agent (CLI with pip/npm install)
Does Vinv require an account or API key?
No—runs entirely locally with no account, API keys, or telemetry.
How does Vinv verify a fix actually works?
Acceptance tests written before the fix are executed; the fix only lands if tests pass. Failing attempts are auto-reverted.
What are the system requirements?
Python, Rust, uv, a coding-agent CLI, and ~500MB for a one-time local embedding model download. Setup takes ~4 minutes.