engineering / agent

jcode

A coding agent harness with a terminal interface, agent memory, multi-agent swarms and many model providers.

Repository owner 1jehuang

Not rated yet
0 comments

Like 0 Save

Published 11 Oct 2026Listing updated 12 Oct 2026

At a glance

Purpose: A coding agent harness with a terminal interface, agent memory, multi-agent swarms and many model providers.
Runs in: Local agent with shell
Operating systems: macOS, Linux, Windows
Needs: Runtime: Not stated; Needs internet: Not stated; Needs accounts or API keys beyond your agent: Not stated
Files and data: Reads files: Yes; Writes files: Yes; Sends data to other services: Not stated
Tool type: Agent
Delivery: From GitHub
Version: Latest on the master branch
Access: Code on GitHub
Compatibility: According to the README, jcode runs on Linux, macOS and Windows. It lists login flows for Claude, OpenAI / ChatGPT / Codex, Google Gemini, GitHub Copilot, Azure OpenAI, Alibaba Cloud Coding Plan, Fireworks, Novita AI, MiniMax, Meta Model API / Muse, LM Studio and Ollama, built-in OpenAI-compatible profiles such as OpenRouter and DeepSeek, and custom OpenAI-compatible and Anthropic-compatible endpoints. The README says jcode reads Claude Code's MCP configuration files and can perform a one-time import of Codex CLI MCP configuration.
External costs: According to the README, jcode works with subscription-backed sign-in flows, so you can use the models you already pay for, and with direct API providers. It names an Alibaba Cloud Coding Plan, and lists Novita AI as using an API key. It says local servers without auth need no API key. The repository is MIT-licensed. The model provider or plan you connect may charge for usage.
Permissions: The README shows downloaded installer scripts executed through bash or PowerShell, Homebrew installation and source builds. The README describes use of subscription-backed sign-ins and user-provided API keys, home-directory configuration, stored provider credentials and a one-time Codex MCP import that can copy secret environment values. The README's agent setup prompt asks the agent to inspect existing credentials, configuration and CLI sessions before choosing a provider. The README describes a persistent background server, scheduled workers, background memory extraction and consolidation, and worker swarms. The README says self-development edits, builds and reloads jcode's own source and binary. The README describes Firefox browser automation, file-changing agents and update downloads. The README also documents an uninstall route that preserves state and a purge route that deletes configuration, authentication, sessions, logs and memory.
Usage terms: MIT License, as stated in the repository. See its LICENSE file for the exact terms.

Code on GitHub

1jehuang/jcode

MIT

Rust20,612 stars2,399 forksLast pushed 12 Oct 2026 UTC

High performance coding agent harness written in rust

Topics: ai ai-agent ai-coding-agent claude cli coding-agent llm mcp

Public GitHub data fetched 13 Oct 2026 UTC · 0 days old

Skill folders in this repository (1)

Folders that contain a SKILL.md file, from GitHub's file list on 13 Oct 2026 UTC.

.jcode/skills/optimization

View on GitHub ↗

What it does

The GitHub description identifies jcode as a coding agent harness written in Rust. The README describes a terminal interface with Mermaid diagrams, side panels, information widgets and custom scrollback. The README describes /update and jcode update for updating the application.

The README describes semantic embeddings, a memory graph, explicit memory tools and session search. The README says swarm agents can work in the same repository, receive file-change notices and message one another. The README says agents can spawn worker swarms.

The README lists subscription-backed sign-in flows, direct API providers and custom OpenAI-compatible and Anthropic-compatible endpoints. The README describes its own MCP files, live reading of Claude Code MCP configuration and a one-time Codex configuration import.

Who it is for

Developers who want a terminal coding agent that uses the model subscriptions or API providers they already have and can run several agents at once.

Example tasks

Sign in to Claude with jcode login --provider claude.
Add a self-hosted OpenAI-compatible endpoint with the jcode provider add command.
Update with /update inside the terminal interface, or with jcode update from a terminal.
Run two or more agents in one repository and let the server pass file-change notices between them.
Load a Markdown file into the side panel or use the panel as a diff viewer.

Getting started

The README gives an install script for macOS and Linux and a PowerShell script for Windows 11. It points to a detailed installation section for Homebrew, source builds and provider setup. To sign in to a provider, it says to run jcode login --provider <profile-id>, and running jcode login with no provider shows the interactive list. For self-hosted endpoints it describes jcode provider add. The README has the exact commands.

Limitations

The README lists Linux x86_64/aarch64, macOS Apple Silicon/Intel, Windows x86_64 with native and WSL2 support, and Termux aarch64/x86_64. The README's Windows installer requires Windows 11 and PowerShell 5.1+; explicit Windows source builds require Git, Rust and Visual Studio 2022 Build Tools with the Desktop development with C++ workload. The README says Termux requires glibc and patchelf. The README says public remote HTTP endpoints are rejected, while localhost and private LAN HTTP endpoints are accepted. The README says MCP support is limited to stdio servers and the built-in browser backend is Firefox. The README warns that weaker models can introduce breaking changes during self-development. The README describes Handterm as a work in progress and notes terminal limits on smooth partial-line scrolling. The repository can change at any time; check its README before you install.

Data practices

According to the README, jcode sends prompts to the provider you sign in to or configure, including local servers. The README says extracted memories are put into a memory graph and that the harness provides session search over previous sessions. It says jcode provider add stores secrets in jcode's private app config directory when requested, and --api-key-env references an existing environment variable instead of storing a key. It warns that environment values imported from a Codex CLI configuration are copied and may contain secrets. The repository also has a TELEMETRY.md file; check it before you install.