A terminal coding agent with built-in LSP, debugger, browser and subagent tools, supporting 60+ model providers.
1jehuang/jcode
A Rust-based coding agent harness built to run many AI coding agents in parallel with low memory use and fast startup.
Open-source alternative to:Claude Code
OpenAI Codex CLI
- Stars
- 21,593
- Forks
- 2,512
- Open issues
- 800
- Last commit
- 2 hours ago
What is it?
jcode is a terminal coding agent written in Rust. Its main goal is running many agents at once on one machine: the project reports about 3.3 MB of extra RAM per additional headless session and a time to first input of about 19 ms in its own benchmarks, which it compares against OpenCode, pi, Claude Code and Codex CLI.
When several agents work in the same repository, a server coordinates them. If one agent edits a file another agent has read, the other is notified, and agents can message each other directly or broadcast to the group. Agents can also spawn their own sub-agents, with the main agent acting as coordinator.
The project supports multiple model providers, including Claude OAuth and OpenAI Codex OAuth. It also includes a self-development mode in which the agent edits, builds and tests its own source, then reloads its binary. Installation is a single curl command on macOS and Linux.
Key features
- 01Runs many headless agent sessions with low per-session RAM use.
- 02Server-managed multi-agent swarms with file-change notifications.
- 03Direct and broadcast messaging between agents in the same repo.
- 04Agents can spawn sub-agents and act as coordinators.
- 05Supports Claude OAuth and OpenAI Codex OAuth among many providers.
- 06Self-development mode that edits, builds and tests its own source.
Use cases
- Running several coding agents in parallel on the same repository.
- Reducing memory pressure when working with many concurrent agent sessions.
- Experimenting with agents that improve their own tooling.
Quick start
- 1
Run the official install script from the README (macOS & Linux)
Things to know
- The README only documents installation for macOS and Linux.
In collections
Topics
FAQ
- What is jcode?
- A Rust-based coding agent harness built to run many AI coding agents in parallel with low memory use and fast startup. jcode is a terminal coding agent written in Rust. Its main goal is running many agents at once on one machine: the project reports about 3.3 MB of extra RAM per additional headless session and a time to first input of about 19 ms in its own benchmarks, which it compares against OpenCode, pi, Claude Code and Codex CLI.
- Is jcode free to use? What license does it have?
- jcode is open source under the MIT license. Check the license terms before using it in commercial products.
- Is jcode still maintained?
- Actively maintained. The last commit to jcode was on Nov 7, 2026.
- What can I use jcode for?
- Running several coding agents in parallel on the same repository. Reducing memory pressure when working with many concurrent agent sessions. Experimenting with agents that improve their own tooling.
- How do I install jcode?
- Run the official install script from the README (macOS & Linux)
- Where can I get jcode?
- The source code and installation instructions are on GitHub: https://github.com/1jehuang/jcode
Summary generated by AI from the project's README. GitHub stats refresh daily.
Related projects
An open-source AI coding agent that runs in your terminal, with companion desktop, web, editor, and chat interfaces.
A terminal coding agent written in Go that edits files, runs commands and keeps local sessions, with you choosing the model and permission level.
A batteries-included Python coding-agent runtime usable from the terminal, an HTTP/WebSocket server, and Telegram.
A CLI-first open-source AI coding agent runtime that works with any model via MCP tools, interactively, piped, as a daemon, or over WebSocket/ACP.
Open-source Rust coding agent for the terminal that edits files, runs commands and works with 40+ model providers, including local models.