jcode
ActiveDescription
A next-generation coding-agent harness built for multi-session workflows, with a focus on extreme performance, low memory, and deep MCP/tool customization.
Key Features
- Extreme footprint - 27.8 MB per single session and 117 MB for 10 sessions, 5-15x lighter than peers
- Sub-millisecond startup - 14 ms to first frame and 49 ms to first input, 200x+ faster than Claude Code
- Multi-session swarm - coordinates multiple agents in the same repo with conflict detection and broadcast messaging
- Semantic memory graph - embeds turns into a vector graph, retrieves relevant context, and consolidates periodically
- Rich provider support - built-in login flows for Claude, OpenAI, Gemini, Copilot, Ollama, LM Studio
- MCP and OpenAI-compatible endpoints - plug self-hosted or third-party APIs via config files
Use Cases
💡 Replace Claude Code or OpenCode when running long-running or parallel sessions with tight memory budgets
💡 Run multiple agents that coordinate on large refactors, test authoring, and code review
💡 Drive end-to-end coding - implement, test, debug - from a single natural-language prompt
💡 Plug your team's MCP tools and OpenAI-compatible LLMs into jcode
💡 Stay in the terminal from prototype through debugging and git commit
Strengths & Limitations
✅ Strengths
- • Actively maintained, recent updates
- • High community interest (19.0k stars)
- • Permissive open-source license (MIT)
Tags
Categories
Quick Start
# macOS / Linux one-liner install
curl -fsSL https://raw.githubusercontent.com/1jehuang/jcode/master/scripts/install.sh | bash
# Sign in with a Claude subscription
jcode login --provider claude
# Or sign in with OpenAI
jcode login --provider openai
# Launch a 3-agent swarm
jcode swarm 3