For the past few months, we've been working on a problem every developer using AI coding tools knows too well:
How do you run multiple AI agents on the same codebase without them creating file conflicts, losing context, or requiring constant manual coordination?
Our answer is Jupiter, a parallel AI orchestration engine written in Rust.
Here's how it works:
You describe a complex task (e.g., "refactor the auth module")
Jupiter analyzes your codebase and creates an execution plan
It launches N Claude CLI workers simultaneously
A Rust-based scheduler manages file locking, dependency ordering, and inter-worker communication
5.You monitor everything from a real-time terminal dashboard
The critical design decision: orchestration logic runs in Rust, not in an LLM. Claude is invoked only for reasoning and writing code. Everything else costs zero tokens.
The result:
- Parallel execution without conflicts
- Cost efficiency (80% Rust, 20% Claude)
- Full audit trail and reproducibility
- Single binary, no external dependencies
We're currently in alpha, with a desktop GUI (Tauri + React) in development.
We're building Jupiter in public with a small community of developers. Join us on Discord to follow the development, test early builds, and shape what
comes next:
hashtag#AI hashtag#DevTools hashtag#Rust hashtag#DeveloperProductivity hashtag#AIAgents hashtag#BuildInPublic
Top comments (0)