◎
Parallel AI Agents
Multiple agents. Many ideas. Only better code survives.
Ship better code. Automatically.
GradeX runs parallel AI agents on your codebase, measures what matters, and keeps only the changes that improve your metrics.
MIT License
Python 3.11+
185 Tests Passing
Open Source
▥
Measure Everything
Benchmarks, tests, and custom metrics you define.
◇
Safe by Design
Built-in gates ensure quality, correctness, and reliability.
⌁
Define what better means
Set up benchmarks, tests, or any metric. GradeX optimizes against numbers, not guesses.
◎
Let AI explore
Agents make many small changes in parallel using isolated git worktrees.
♕
Keep the winners
Only changes that beat your baseline and pass your gates are kept.
Built with
PythonGitFastAPISQLiteTyperRich
Model providers
Free by default. Bring your own model when you need more.
GradeX supports hosted and local providers, with Groq as the recommended free cloud option.
Groq
Use llama-3.3-70b-versatile with a free API key from console.groq.com.
Anthropic
Run Claude models for deeper code reasoning and large refactors.
OpenAI
Use GPT models through the standard OpenAI-compatible provider path.
Ollama
Run fully local models with no account, no cloud calls, and no API bill.
Host integrations
Install into the coding tools your team already uses.
Cursor
Installs project rules into .cursor/rules/ so agents know how to run GradeX.
gradex install cursor
Claude Code
Installs native skill files for discover and optimize workflows.
gradex install claude-code
CLI-first
Everything also works directly from the terminal without an agent host.
gradex --help
Install in 30 seconds
Start with one command.
No waitlist, no hosted platform required. Install GradeX from PyPI and run it locally.
pip install gradex
gradex install cursor
gradex discover "make the parser faster" --provider groq --api-key <key>
Ready to improve your code?
Get started in minutes. No credit card required.