[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"project-223":3},{"id":4,"name":5,"fullName":6,"owner":7,"repo":5,"description":8,"homepage":9,"htmlUrl":10,"language":11,"languages":9,"totalLinesOfCode":9,"stars":12,"forks":13,"watchers":14,"openIssues":15,"contributorsCount":9,"subscribersCount":16,"size":16,"stars1d":17,"stars7d":18,"stars30d":19,"stars90d":16,"forks30d":16,"starsTrendScore":20,"compositeScore":21,"rankGlobal":9,"rankLanguage":9,"license":9,"archived":22,"fork":22,"defaultBranch":23,"hasWiki":22,"hasPages":22,"topics":24,"createdAt":9,"pushedAt":9,"updatedAt":35,"readmeContent":36,"aiSummary":37,"trendingCount":16,"starSnapshotCount":16,"syncStatus":38,"lastSyncTime":39,"discoverSource":40},223,"jcode","1jehuang\u002Fjcode","1jehuang","Coding Agent Harness",null,"https:\u002F\u002Fgithub.com\u002F1jehuang\u002Fjcode","Rust",7008,786,38,92,0,25,137,1222,119,39.69,false,"main",[25,26,27,28,29,30,31,32,33,34],"ai","claude","cli","coding-agent","llm","mcp","openai","rust","terminal","tui","2026-06-12 02:00:10","\u003Cdiv align=\"center\">\n\n# jcode\n\n[![Latest Release](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fv\u002Frelease\u002F1jehuang\u002Fjcode?style=flat-square)](https:\u002F\u002Fgithub.com\u002F1jehuang\u002Fjcode\u002Freleases)\n[![License](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Flicense\u002F1jehuang\u002Fjcode?style=flat-square)](LICENSE)\n[![Platforms](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fplatforms-Linux%20%7C%20macOS%20%7C%20Windows-blue?style=flat-square)](https:\u002F\u002Fgithub.com\u002F1jehuang\u002Fjcode\u002Freleases)\n[![Commit Activity](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fcommit-activity\u002Fm\u002F1jehuang\u002Fjcode?style=flat-square)](https:\u002F\u002Fgithub.com\u002F1jehuang\u002Fjcode\u002Fcommits\u002Fmaster)\n[![GitHub Stars](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002F1jehuang\u002Fjcode?style=flat-square)](https:\u002F\u002Fgithub.com\u002F1jehuang\u002Fjcode\u002Fstargazers)\n\nThe next generation coding agent harness to raise the skill ceiling. \u003Cbr>\nBuilt for multi-session workflows, infinite customizability, and performance. \n\n\u003Cbr>\n\n\u003Ca href=\"https:\u002F\u002Fgithub.com\u002F1jehuang\u002Fjcode\u002Freleases\u002Fdownload\u002Freadme-assets\u002Fjcode-memory-demo.mp4\">\n  \u003Cimg src=\"https:\u002F\u002Fgithub.com\u002F1jehuang\u002Fjcode\u002Freleases\u002Fdownload\u002Freadme-assets\u002Fjcode-memory-demo.webp\" alt=\"jcode memory demonstration\" width=\"800\">\n\u003C\u002Fa>\n\n\u003Cbr>\n\n[Features](#features) · [Install](#installation) · [Quick Start](#quick-start) · [Further Reading](#further-reading) · [Contributing](CONTRIBUTING.md)\n\n\u003C\u002Fdiv>\n\n---\n\n\u003Cdiv align=\"center\">\n\n## Installation\n\n\u003C\u002Fdiv>\n\n```bash\n# macOS & Linux\ncurl -fsSL https:\u002F\u002Fraw.githubusercontent.com\u002F1jehuang\u002Fjcode\u002Fmaster\u002Fscripts\u002Finstall.sh | bash\n```\n\nNeed Windows, Homebrew, source builds, provider setup, or tell your agent to set it up for you?\n[Jump to detailed installation](#detailed-installation).\n\n---\n\n\n\u003Cdiv align=\"center\">\n\n## Performance & Resource Efficiency\n\n\u003C\u002Fdiv>\n\njcode is built to be as performant and resource efficient as possible. Every metric is optimized to the bone, which is important for scaling multi-session workflows. Here we sample a few metrics to show the difference: RAM usage and boot up.\n\n### RAM comparison\n\n\u003Cdiv align=\"center\">\n\n\u003Ctable>\n  \u003Ctr>\n    \u003Ctd valign=\"top\" align=\"center\" width=\"50%\">\n      \u003Cstrong>1 active session\u003C\u002Fstrong>\n      \u003Ctable>\n        \u003Cthead>\n          \u003Ctr>\n            \u003Cth>Tool\u003C\u002Fth>\n            \u003Cth>PSS\u003C\u002Fth>\n            \u003Cth>Comparison\u003C\u002Fth>\n          \u003C\u002Ftr>\n        \u003C\u002Fthead>\n        \u003Ctbody>\n          \u003Ctr>\n            \u003Ctd>\u003Cstrong>jcode (local embedding off)\u003C\u002Fstrong>\u003C\u002Ftd>\n            \u003Ctd align=\"right\">\u003Cstrong>27.8 MB\u003C\u002Fstrong>\u003C\u002Ftd>\n            \u003Ctd align=\"right\">baseline\u003C\u002Ftd>\n          \u003C\u002Ftr>\n          \u003Ctr>\n            \u003Ctd>\u003Cstrong>jcode\u003C\u002Fstrong>\u003C\u002Ftd>\n            \u003Ctd align=\"right\">\u003Cstrong>167.1 MB\u003C\u002Fstrong>\u003C\u002Ftd>\n            \u003Ctd align=\"right\">\u003Cstrong>6.0× more RAM\u003C\u002Fstrong>\u003C\u002Ftd>\n          \u003C\u002Ftr>\n          \u003Ctr>\n            \u003Ctd>\u003Cstrong>pi\u003C\u002Fstrong>\u003C\u002Ftd>\n            \u003Ctd align=\"right\">\u003Cstrong>144.4 MB\u003C\u002Fstrong>\u003C\u002Ftd>\n            \u003Ctd align=\"right\">\u003Cstrong>5.2× more RAM\u003C\u002Fstrong>\u003C\u002Ftd>\n          \u003C\u002Ftr>\n          \u003Ctr>\n            \u003Ctd>\u003Cstrong>Codex CLI\u003C\u002Fstrong>\u003C\u002Ftd>\n            \u003Ctd align=\"right\">\u003Cstrong>140.0 MB\u003C\u002Fstrong>\u003C\u002Ftd>\n            \u003Ctd align=\"right\">\u003Cstrong>5.0× more RAM\u003C\u002Fstrong>\u003C\u002Ftd>\n          \u003C\u002Ftr>\n          \u003Ctr>\n            \u003Ctd>\u003Cstrong>OpenCode\u003C\u002Fstrong>\u003C\u002Ftd>\n            \u003Ctd align=\"right\">\u003Cstrong>371.5 MB\u003C\u002Fstrong>\u003C\u002Ftd>\n            \u003Ctd align=\"right\">\u003Cstrong>13.4× more RAM\u003C\u002Fstrong>\u003C\u002Ftd>\n          \u003C\u002Ftr>\n          \u003Ctr>\n            \u003Ctd>\u003Cstrong>GitHub Copilot CLI\u003C\u002Fstrong>\u003C\u002Ftd>\n            \u003Ctd align=\"right\">\u003Cstrong>333.3 MB\u003C\u002Fstrong>\u003C\u002Ftd>\n            \u003Ctd align=\"right\">\u003Cstrong>12.0× more RAM\u003C\u002Fstrong>\u003C\u002Ftd>\n          \u003C\u002Ftr>\n          \u003Ctr>\n            \u003Ctd>\u003Cstrong>Cursor Agent\u003C\u002Fstrong>\u003C\u002Ftd>\n            \u003Ctd align=\"right\">\u003Cstrong>214.9 MB\u003C\u002Fstrong>\u003C\u002Ftd>\n            \u003Ctd align=\"right\">\u003Cstrong>7.7× more RAM\u003C\u002Fstrong>\u003C\u002Ftd>\n          \u003C\u002Ftr>\n          \u003Ctr>\n            \u003Ctd>\u003Cstrong>Claude Code\u003C\u002Fstrong>\u003C\u002Ftd>\n            \u003Ctd align=\"right\">\u003Cstrong>386.6 MB\u003C\u002Fstrong>\u003C\u002Ftd>\n            \u003Ctd align=\"right\">\u003Cstrong>13.9× more RAM\u003C\u002Fstrong>\u003C\u002Ftd>\n          \u003C\u002Ftr>\n        \u003C\u002Ftbody>\n      \u003C\u002Ftable>\n    \u003C\u002Ftd>\n    \u003Ctd width=\"24\">\u003C\u002Ftd>\n    \u003Ctd valign=\"top\" align=\"center\" width=\"50%\">\n      \u003Cstrong>10 active sessions\u003C\u002Fstrong>\n      \u003Ctable>\n        \u003Cthead>\n          \u003Ctr>\n            \u003Cth>Tool\u003C\u002Fth>\n            \u003Cth>PSS\u003C\u002Fth>\n            \u003Cth>Comparison\u003C\u002Fth>\n          \u003C\u002Ftr>\n        \u003C\u002Fthead>\n        \u003Ctbody>\n          \u003Ctr>\n            \u003Ctd>\u003Cstrong>jcode (local embedding off)\u003C\u002Fstrong>\u003C\u002Ftd>\n            \u003Ctd align=\"right\">\u003Cstrong>117.0 MB\u003C\u002Fstrong>\u003C\u002Ftd>\n            \u003Ctd align=\"right\">baseline\u003C\u002Ftd>\n          \u003C\u002Ftr>\n          \u003Ctr>\n            \u003Ctd>\u003Cstrong>jcode\u003C\u002Fstrong>\u003C\u002Ftd>\n            \u003Ctd align=\"right\">\u003Cstrong>260.8 MB\u003C\u002Fstrong>\u003C\u002Ftd>\n            \u003Ctd align=\"right\">\u003Cstrong>2.2× more RAM\u003C\u002Fstrong>\u003C\u002Ftd>\n          \u003C\u002Ftr>\n          \u003Ctr>\n            \u003Ctd>\u003Cstrong>pi\u003C\u002Fstrong>\u003C\u002Ftd>\n            \u003Ctd align=\"right\">\u003Cstrong>833.0 MB\u003C\u002Fstrong>\u003C\u002Ftd>\n            \u003Ctd align=\"right\">\u003Cstrong>7.1× more RAM\u003C\u002Fstrong>\u003C\u002Ftd>\n          \u003C\u002Ftr>\n          \u003Ctr>\n            \u003Ctd>\u003Cstrong>Codex CLI\u003C\u002Fstrong>\u003C\u002Ftd>\n            \u003Ctd align=\"right\">\u003Cstrong>334.8 MB\u003C\u002Fstrong>\u003C\u002Ftd>\n            \u003Ctd align=\"right\">\u003Cstrong>2.9× more RAM\u003C\u002Fstrong>\u003C\u002Ftd>\n          \u003C\u002Ftr>\n          \u003Ctr>\n            \u003Ctd>\u003Cstrong>OpenCode\u003C\u002Fstrong>\u003C\u002Ftd>\n            \u003Ctd align=\"right\">\u003Cstrong>3237.2 MB\u003C\u002Fstrong>\u003C\u002Ftd>\n            \u003Ctd align=\"right\">\u003Cstrong>27.7× more RAM\u003C\u002Fstrong>\u003C\u002Ftd>\n          \u003C\u002Ftr>\n          \u003Ctr>\n            \u003Ctd>\u003Cstrong>GitHub Copilot CLI\u003C\u002Fstrong>\u003C\u002Ftd>\n            \u003Ctd align=\"right\">\u003Cstrong>1756.5 MB\u003C\u002Fstrong>\u003C\u002Ftd>\n            \u003Ctd align=\"right\">\u003Cstrong>15.0× more RAM\u003C\u002Fstrong>\u003C\u002Ftd>\n          \u003C\u002Ftr>\n          \u003Ctr>\n            \u003Ctd>\u003Cstrong>Cursor Agent\u003C\u002Fstrong>\u003C\u002Ftd>\n            \u003Ctd align=\"right\">\u003Cstrong>1632.4 MB\u003C\u002Fstrong>\u003C\u002Ftd>\n            \u003Ctd align=\"right\">\u003Cstrong>14.0× more RAM\u003C\u002Fstrong>\u003C\u002Ftd>\n          \u003C\u002Ftr>\n          \u003Ctr>\n            \u003Ctd>\u003Cstrong>Claude Code\u003C\u002Fstrong>\u003C\u002Ftd>\n            \u003Ctd align=\"right\">\u003Cstrong>2300.6 MB\u003C\u002Fstrong>\u003C\u002Ftd>\n            \u003Ctd align=\"right\">\u003Cstrong>19.7× more RAM\u003C\u002Fstrong>\u003C\u002Ftd>\n          \u003C\u002Ftr>\n        \u003C\u002Ftbody>\n      \u003C\u002Ftable>\n    \u003C\u002Ftd>\n  \u003C\u002Ftr>\n\u003C\u002Ftable>\n\n\u003C\u002Fdiv>\n\n### Time to first frame\n\n\u003Cdiv align=\"center\">\n\n| Tool | Time to first frame | Range | Comparison |\n|---|---:|---:|---:|\n| **jcode** | **14.0 ms** | 10.1–19.3 ms | baseline |\n| **pi** | **590.7 ms** | 369.6–934.8 ms | **42.2× slower** |\n| **Codex CLI** | **882.8 ms** | 742.3–1640.9 ms | **63.1× slower** |\n| **OpenCode** | **1035.9 ms** | 922.5–1104.4 ms | **74.0× slower** |\n| **GitHub Copilot CLI** | **1518.6 ms** | 1357.4–1826.8 ms | **108.5× slower** |\n| **Cursor Agent** | **1949.7 ms** | 1711.0–2104.8 ms | **139.3× slower** |\n| **Claude Code** | **3436.9 ms** | 2032.7–8927.2 ms | **245.5× slower** |\n\n\u003C\u002Fdiv>\n\nMeasured on this Linux machine across 10 interactive PTY launches.\n\n### Time to first input\n(time until typed probe text appears on the rendered screen.)\n\u003Cdiv align=\"center\">\n\n| Tool | Time to first input | Range | Comparison |\n|---|---:|---:|---:|\n| **jcode** | **48.7 ms** | 30.3–62.7 ms | baseline |\n| **pi** | **596.4 ms** | 373.9–955.2 ms | **12.2× slower** |\n| **Codex CLI** | **905.8 ms** | 760.1–1675.7 ms | **18.6× slower** |\n| **OpenCode** | **1047.9 ms** | 931.1–1116.9 ms | **21.5× slower** |\n| **GitHub Copilot CLI** | **1583.4 ms** | 1422.8–1880.0 ms | **32.5× slower** |\n| **Cursor Agent** | **1978.7 ms** | 1727.3–2130.0 ms | **40.6× slower** |\n| **Claude Code** | **3512.8 ms** | 2137.4–9002.0 ms | **72.2× slower** |\n\n\u003C\u002Fdiv>\n\nMeasured on this Linux machine across 10 interactive PTY launches.\n\n### Additional clients \u002F memory scaling\n\n\u003Cdiv align=\"center\">\n\n| Tool | Extra PSS per added session | Comparison |\n|---|---:|---:|\n| **jcode (local embedding off)** | **~9.9 MB** | baseline |\n| **jcode** | **~10.4 MB** | **1.1× more RAM** |\n| **pi** | **~76.5 MB** | **7.7× more RAM** |\n| **Codex CLI** | **~21.6 MB** | **2.2× more RAM** |\n| **OpenCode** | **~318.4 MB** | **32.2× more RAM** |\n| **GitHub Copilot CLI** | **~158.1 MB** | **16.0× more RAM** |\n| **Cursor Agent** | **~157.5 MB** | **15.9× more RAM** |\n| **Claude Code** | **~212.7 MB** | **21.5× more RAM** |\n\n\u003C\u002Fdiv>\nversions tested for this corrected memory rerun:\n\n- `jcode v0.9.1888-dev (be386f2)`\n- `pi 0.62.0`\n- `codex-cli 0.120.0`\n- `opencode 1.0.203`\n- `GitHub Copilot CLI 1.0.24` for the 1-session rerun, `GitHub Copilot CLI 1.0.27` for the 10-session rerun\n- `Cursor Agent 2026.04.08-a41fba1`\n- `Claude Code 2.1.86 (Claude Code)`\n\n\u003Cdiv align=\"center\">\n\n  \u003Ca href=\"https:\u002F\u002Fgithub.com\u002F1jehuang\u002Fjcode\u002Freleases\u002Fdownload\u002Freadme-assets\u002Fjcode-performance-demo.mp4\">\n    \u003Cimg src=\"https:\u002F\u002Fgithub.com\u002F1jehuang\u002Fjcode\u002Freleases\u002Fdownload\u002Freadme-assets\u002Fjcode-performance-demo.webp\" alt=\"jcode performance demonstration\" width=\"900\">\n  \u003C\u002Fa>\n\n  \u003Cp>\u003Cem>jcode performance demonstration\u003C\u002Fem>\u003C\u002Fp>\n\n\u003C\u002Fdiv>\n\n\n---\n\n## Memory (Agent memory)\n\nJcode embeds each turn\u002Fresponse as a semantic vector. Every turn does queries a graph of memories to efficiently find related memory entries via a cosine similarity check. The embedding hits are fed into the conversation, or optionally uses a memory sideagent which verifies the memories are relevant, and potentially does more work for information retreival before injecting into the conversation. This results in a human like memory system which allows the agent to automatically recall relevant information to the conversation without actively calling memory tools or being a token burner. \not \nTo have memories which are retrieved, they must also be extracted and stored. Every so often (semantic drift, K turns since last extraction, session end, etc), memories are extracted via a memory sideagent, and put into the memory graph. \n\nThe harness also provides explicit memory tools to allow the agent to actively search or store the memory without relying on a passive background process. The harness also provides session search for traditional RAG on previous sessions. \n\nMemories are automatically consolidated every so often via the ambient mode. This reorganizes, checks for staleness and conflicts, etc\n\n\u003Cdiv align=\"center\">\n\n  \u003Ca href=\"https:\u002F\u002Fgithub.com\u002F1jehuang\u002Fjcode\u002Freleases\u002Fdownload\u002Freadme-assets\u002Fjcode-memory-demo.mp4\">\n    \u003Cimg src=\"https:\u002F\u002Fgithub.com\u002F1jehuang\u002Fjcode\u002Freleases\u002Fdownload\u002Freadme-assets\u002Fjcode-memory-demo.webp\" alt=\"jcode memory demonstration\" width=\"900\">\n  \u003C\u002Fa>\n\n  \u003Cp>\u003Cem>jcode memory demonstration\u003C\u002Fem>\u003C\u002Fp>\n\n\u003C\u002Fdiv>\n\n\u003C!-- Memory demo media is hosted in the readme-assets release. -->\n\n---\n\n## UI: Side panels, Diagrams, Info Widgets, rendering, scrolling, alignment\n\nThe side panel is a place for auxiliary information. Tell your jcode agent to load a file into the side panel and see it update in real time, or tell your agent to write directly to the side panel, or use it as a diff viewer. The side panel (and chat) is able to render mermaid diagrams inline. \n\u003Cimg width=\"2877\" height=\"1762\" alt=\"image\" src=\"https:\u002F\u002Fgithub.com\u002Fuser-attachments\u002Fassets\u002F6c7bec81-ef3f-434d-8a7b-d55f8a54e5cf\" \u002F>\n\nTo make this possible, I created a new mermaid rendering library to render diagrams 1800x faster. It has no browser or Typescript dependency. See https:\u002F\u002Fgithub.com\u002F1jehuang\u002Fmermaid-rs-renderer\n\nTo show you important information without taking space away from the screen that could be used for responses, I developed info widgets. Info widgets will only ever take up the negative space on the screen to show you information, and will get out of the way if there isn't any. \n\nJcode can render at over a thousand fps. Your monitor will not have the refresh rate to show you, but this means you will not have silly flicker problems. \n\nThe custom scrollback implementation of jcode allows it to do much more than a native scrollback. However, it is a terminal-level limitation that I cannot have smooth, partial line scrolling with a custom scrollback. To fix this, I made my own terminal. Handterm https:\u002F\u002Fgithub.com\u002F1jehuang\u002Fhandterm implements a native scroll api, and also happens to be very effiecent. This is a work in progress. Scrolling is still well implemented for normal terminals.\n\nJcode is left-aligned by default. You can switch to centered mode with the `Alt+C` hotkey, with the `\u002Falignment` command, or in the config.\n\n---\n\n## Swarm\n\nSpawn two or more agents in the same repo, and they will automatically be managed by the server to allow native collaboration. When agent A edits a file that agent B has read (code shifting under its feet), the server notifies agent B. Agent B can ignore it if it is not relevant, or it can check the diff to make sure that it doesn't conflict. Each agent has messaging abilities, capable of DMing just one agent, broadcasting to all other agents hosted by the server, or just agents working in that repo. This allows you to spawn multiple sessions in the same repo, and have all conflicts automatically resolved.\n\n\u003Cdiv align=\"center\">\n\n  \u003Ca href=\"https:\u002F\u002Fgithub.com\u002F1jehuang\u002Fjcode\u002Freleases\u002Fdownload\u002Freadme-assets\u002Fswarm-demo.mp4\">\n    \u003Cimg src=\"https:\u002F\u002Fgithub.com\u002F1jehuang\u002Fjcode\u002Freleases\u002Fdownload\u002Freadme-assets\u002Fjcode-swarm-demonstration.webp\" alt=\"jcode swarm demonstration\" width=\"900\">\n  \u003C\u002Fa>\n\n  \u003Cp>\u003Cem>jcode swarm demonstration\u003C\u002Fem>\u003C\u002Fp>\n\n\u003C\u002Fdiv>\n\nAgents are also able to spawn their own swarms autonomously. They have a swarm tool which allows them to spawn in their own teamates to accomplish tasks in parallel. Doing so turns the main agent into a coordinator and the spawned agents into workers. Groups of agents, their messaging channels, their completion statuses, etc are all automatically managed. This can be done headlessly or headed.\n\n---\n\n## OAuth and Providers\n\njcode works with subscription-backed OAuth flows and many provider integrations, so you can use the models you already pay for and still fall back to direct API providers when needed.\n\n### Supported built-in login flows\n\n- **Claude** (`jcode login --provider claude`)\n- **OpenAI \u002F ChatGPT \u002F Codex** (`jcode login --provider openai`)\n- **Google Gemini** (`jcode login --provider gemini`)\n- **GitHub Copilot** (`jcode login --provider copilot`)\n- **Azure OpenAI** (`jcode login --provider azure`)\n- **Alibaba Cloud Coding Plan** (`jcode login --provider alibaba-coding-plan`)\n- **Fireworks** (`jcode login --provider fireworks`)\n- **MiniMax** (`jcode login --provider minimax`)\n- **LM Studio** (`jcode login --provider lmstudio`)\n- **Ollama** (`jcode login --provider ollama`)\n- **Custom OpenAI-compatible endpoint** (`jcode login --provider openai-compatible`)\n\nFor custom OpenAI-compatible endpoints, jcode now prompts for the API base and supports local localhost servers without requiring an API key.\n\n### Config-file setup for self-hosted endpoints and MCP\n\nIf you prefer to configure things by editing files instead of using the login UI, jcode supports both a custom OpenAI-compatible endpoint config and MCP config files.\n\n#### Self-hosted OpenAI-compatible endpoints, including vLLM\n\nFor agents and scripts, the preferred path is the one-shot provider profile command. It writes a named profile to `~\u002F.jcode\u002Fconfig.toml`, stores secrets in jcode's private app config directory when requested, and prints exact run\u002Fvalidation commands:\n\n```bash\n# Secret-safe setup for a hosted OpenAI-compatible API.\nprintf '%s' \"$MY_API_KEY\" | jcode provider add my-api \\\n  --base-url https:\u002F\u002Fllm.example.com\u002Fv1 \\\n  --model my-model-id \\\n  --api-key-stdin \\\n  --set-default \\\n  --json\n\n# Smoke test the profile.\njcode --provider-profile my-api auth-test --prompt 'Reply exactly JCODE_PROVIDER_SETUP_OK'\n\n# Use it directly.\njcode --provider-profile my-api run 'hello'\n```\n\nFor local servers that do not require auth:\n\n```bash\njcode provider add local-vllm \\\n  --base-url http:\u002F\u002Flocalhost:8000\u002Fv1 \\\n  --model Qwen\u002FQwen3-Coder-30B-A3B-Instruct \\\n  --no-api-key \\\n  --set-default\n```\n\nUseful flags:\n\n- `--api-key-env NAME`: reference an existing environment variable instead of storing a key.\n- `--api-key-stdin`: read and store a key without putting it in shell history.\n- `--context-window TOKENS`: persist the model context window for model selection and routing.\n- `--overwrite`: replace an existing profile of the same name.\n- `--model-catalog`: use the endpoint's `\u002Fmodels` response in addition to configured models.\n\nThe generated profile can also be edited manually in `~\u002F.jcode\u002Fconfig.toml`:\n\n```toml\n[provider]\ndefault_provider = \"my-api\"\ndefault_model = \"my-model-id\"\n\n[providers.my-api]\ntype = \"openai-compatible\"\nbase_url = \"https:\u002F\u002Fllm.example.com\u002Fv1\"\napi_key_env = \"JCODE_PROVIDER_MY_API_API_KEY\"\nenv_file = \"provider-my-api.env\"\ndefault_model = \"my-model-id\"\n\n[[providers.my-api.models]]\nid = \"my-model-id\"\ncontext_window = 128000\n```\n\nThe custom OpenAI-compatible provider reads overrides from environment variables or from an env file in jcode's app config directory. On Linux this is usually `~\u002F.config\u002Fjcode\u002F`, so the default file is usually:\n\n```text\n~\u002F.config\u002Fjcode\u002Fopenai-compatible.env\n```\n\nExample for a local or LAN vLLM server:\n\n```bash\nJCODE_OPENAI_COMPAT_API_BASE=http:\u002F\u002F192.168.1.50:8000\u002Fv1\nJCODE_OPENAI_COMPAT_DEFAULT_MODEL=Qwen\u002FQwen3-Coder-30B-A3B-Instruct\n# Optional if your server expects auth\nOPENAI_COMPAT_API_KEY=your-token-here\n```\n\nNotes:\n\n- `jcode login --provider openai-compatible` can create or update this for you.\n- Plain `http:\u002F\u002F` is accepted for `localhost` and private LAN IPs. Public remote HTTP is still rejected.\n- HTTPS endpoints work as usual.\n\n#### MCP config files\n\nMCP config is separate from `config.toml`.\n\nPrimary config files:\n\n- `~\u002F.jcode\u002Fmcp.json` for global MCP servers\n- `.jcode\u002Fmcp.json` for project-local MCP servers\n\nCompatibility fallback:\n\n- `.claude\u002Fmcp.json`\n\nExample MCP config:\n\n```json\n{\n  \"servers\": {\n    \"filesystem\": {\n      \"command\": \"\u002Fpath\u002Fto\u002Fmcp-server\",\n      \"args\": [\"--root\", \"\u002Fworkspace\"],\n      \"env\": {},\n      \"shared\": true\n    }\n  }\n}\n```\n\nOn first run, jcode also tries to import MCP servers from `~\u002F.claude\u002Fmcp.json` and `~\u002F.codex\u002Fconfig.toml` if `~\u002F.jcode\u002Fmcp.json` does not exist yet.\n\nFor headless or SSH sessions, OAuth-style providers support `jcode login --provider \u003Cprovider> --no-browser` (alias: `--headless`) so jcode prints the auth URL\u002FQR and falls back to manual code or callback paste instead of trying to launch a local browser.\n\nFor more scriptable remote flows, `claude`, `openai`, `gemini`, and `antigravity` also support a two-step pattern:\n\n```bash\n# Step 1: print a resumable auth URL\njcode login --provider openai --print-auth-url --json\n\n# Step 2: complete later with the callback URL or auth code\njcode login --provider openai --callback-url 'http:\u002F\u002Flocalhost:1455\u002Fauth\u002Fcallback?...'\njcode login --provider gemini --auth-code '...'\n```\n\nAdditional scriptable cases:\n\n```bash\n# Copilot device flow: print URL + user code, then complete later\njcode login --provider copilot --print-auth-url --json\njcode login --provider copilot --complete\n\n# Gmail\u002FGoogle OAuth after credentials are already configured\njcode login --provider google --print-auth-url --google-access-tier readonly\njcode login --provider google --callback-url 'http:\u002F\u002F127.0.0.1:8456?...'\n```\n\nPending scriptable login state is stored under `~\u002F.jcode\u002Fpending-login\u002F`, automatically expires, and stale entries are cleaned up when new scriptable logins start or resume.\n\nFor the built-in OpenAI login flow, jcode opens a local callback on\n`http:\u002F\u002Flocalhost:1455\u002Fauth\u002Fcallback` by default.\n\n\u003Cimg width=\"2877\" height=\"1762\" alt=\"Screenshot from 2026-04-02 14-28-51\" src=\"https:\u002F\u002Fgithub.com\u002Fuser-attachments\u002Fassets\u002F530684c0-9d12-4363-aa0e-1b39a0d4e1be\" \u002F>\nThe above image is the first page of provider logins\n\n### Supported provider\n\n- **Native \u002F first-party style providers:** `claude`, `openai`, `copilot`, `gemini`, `azure`, `alibaba-coding-plan`\n- **Aggregator \u002F compatibility providers:** `openrouter`, `openai-compatible`\n- **Additional provider integrations:** `opencode`, `opencode-go`, `zai` \u002F `kimi`, `302ai`, `baseten`, `cortecs`, `deepseek`, `firmware`, `huggingface`, `moonshotai`, `nebius`, `scaleway`, `stackit`, `groq`, `mistral`, `perplexity`, `togetherai`, `deepinfra`, `fireworks`, `minimax`, `xai`, `lmstudio`, `ollama`, `chutes`, `cerebras`, `cursor`, `antigravity`, `google`\n\nJcode also supports easy multi-account switching. Ran out of tokens on your first ChatGPT Pro subscription? \u002Faccount and quickly switch to your second. \n\n---\n\n## Customizability \u002F Self-Dev\n\nJcode is inventing a new form of customizability. One that doesn't limit you to what a plugin or extension can do. Tell your jcode agent to enter self dev mode, and it will start modifying its own source code. Jcode is optimized to iterate on itself. There is significant infrastructure around self developement, which allows it to edit, build, and test its own source code, then reload its own binary and continue work in your (potentially many) sessions, fully automatically. \n\nIt is reccomended that you use a frontier model for this. The jcode codebase is not a simple one, and weaker models can make subtle, breaking changes. GPT 5.5 or the latest available frontier model works well.\n\n\u003C!-- Add self-dev demo thumbnail\u002Fvideo and fuller writeup here. -->\n\n---\n\n## Misc.\n\nThe devil is in the details. There are many undocumented optimizations and niceties that jcode implements. Some examples: \n\nAnthropic's Claude cache goes cold after 5 minutes. If you initiate Claude after these 5 minutes, you have a cache miss, potentially costing you lots of tokens. The ui warns you when the cache went cold, and notfies you if there was an unexpected cache miss. \n\njcode comes with instructions on how to set up Firefox Agent Bridge. Ask you agent to set it up, and then you will have browser automation in jcode as well. \n\nAgent grep is a grep tool I made for the jcode agent. It adds file strucuture information (ie the list of functions, their displacement, etc) to the grep return, so that the agent can infer more of what the file doesn without actually reading the file. It also implements a harness-level integration that adaptively truncates returns based on what the agent has already seen. This saves on context a lot. \n\nInputs are by default interleaved with the working agent. It sends the input as soon as it safely can without breaking the KV cache. Submit with shift enter instead, and it will send a queue send, and wait for the agent to fully finish its turn before sending.\n\nResume sessions from different harnesses. Claude code broke on you? Resume the session from jcode and continue where you left off. Session resume is supported for codex, claude code, opencode, and pi. \n\n\u003Cimg width=\"2877\" height=\"1762\" alt=\"Screenshot from 2026-04-11 16-28-52\" src=\"https:\u002F\u002Fgithub.com\u002Fuser-attachments\u002Fassets\u002Fc2b383cf-2531-4217-85ae-6a863354dc97\" \u002F>\nimage of \u002FResume for codex sessions\n\n\nSkills are not all loaded on startup. The conversation is embedded as a semantic vector, and will automatically inject a skill if there is an embedding hit similar to memories. The agent has a skill tool for you to manually activate a skill at anytime. You may also activate via slash commands. \n\n---\n\n## iOS Application \u002F Native OpenClaw\n\nA native iOS application version of jcode is coming soon. This will allow you to work with jcode on your personal machine's environment from your phone, via Tailscale. Openclaw like features will be bundled with this iOS application. \n\n---\n\n## Other planned features\n\nAgents dont like to commit in dirty git state with active changes. Git was clearly not built for multi-agent workflows, and git worktrees is not a good solution. Given this, I believe that is an opporunity for a new git like primitive to be born. \n\nBuild speed improvements: An incremental debug cargo build with cache enabled takes about 1 minute on my machine. The goal is 5-20 seconds. Refactors and crates seams should be able to make this happen. \n\n\u003C!-- Add iOS \u002F native OpenClaw preview and fuller writeup here. -->\n\n---\n\n\u003Cdiv align=\"center\">\n\n## Quick Start\n\n\u003C\u002Fdiv>\n\n```bash\n# Launch the TUI\njcode\n\n# Run a single command non-interactively\njcode run \"say hello\"\n\n# Resume a previous session by memorable name\njcode --resume fox\n\n# Run as a persistent background server, then attach more clients\njcode serve\njcode connect\n\n# Send voice input from your configured STT command\njcode dictate\n```\n\njcode supports interactive TUI use, non-interactive runs, persistent server\u002Fclient workflows,\nand hotkey-friendly dictation without requiring a bundled speech-to-text stack.\n\n\u003Cdiv align=\"center\">\n\n  \u003Ca href=\"https:\u002F\u002Fgithub.com\u002F1jehuang\u002Fjcode\u002Freleases\u002Fdownload\u002Freadme-assets\u002Fworkflow.mp4\">\n    \u003Cimg src=\"https:\u002F\u002Fgithub.com\u002F1jehuang\u002Fjcode\u002Freleases\u002Fdownload\u002Freadme-assets\u002Fjcode-workflow-demonstration.webp\" alt=\"jcode workflow demonstration\" width=\"900\">\n  \u003C\u002Fa>\n\n  \u003Cp>\u003Cem>jcode workflow demonstration\u003C\u002Fem>\u003C\u002Fp>\n\n\u003C\u002Fdiv>\n\n---\n\n## Browser Automation\n\njcode includes a first-class built-in `browser` tool for browser control inside agent sessions.\n\nCurrent built-in backend:\n- Firefox via Firefox Agent Bridge\n\nCurrent built-in tool actions include:\n- `status`\n- `setup`\n- `open`\n- `snapshot`\n- `get_content`\n- `interactables`\n- `click`\n- `type`\n- `fill_form`\n- `select`\n- `wait`\n- `screenshot`\n- `eval`\n- `scroll`\n- `upload`\n- `press`\n\nQuick setup:\n\n```bash\njcode browser status\njcode browser setup\n```\n\nOnce setup is complete, the model can use the built-in `browser` tool directly. The UI also summarizes browser tool calls compactly, for example opening a URL, clicking a selector, or typing into a field without echoing sensitive typed text.\n\nNotes:\n- the provider\u002Ftool architecture is in place for additional backends\n- Firefox is the wired built-in backend today\n- Chrome bridge \u002F remote debugging style providers can be added on top of the same browser tool later\n\n---\n\n## Further Reading\n\n- [Ambient Mode \u002F OpenClaw](docs\u002FAMBIENT_MODE.md)\n- [Browser Provider Protocol](docs\u002FBROWSER_PROVIDER_PROTOCOL.md)\n- [Memory Architecture](docs\u002FMEMORY_ARCHITECTURE.md)\n- [Swarm Architecture](docs\u002FSWARM_ARCHITECTURE.md)\n- [Server Architecture](docs\u002FSERVER_ARCHITECTURE.md)\n- [iOS Client Notes](docs\u002FIOS_CLIENT.md)\n- [Safety System](docs\u002FSAFETY_SYSTEM.md)\n- [Windows Notes](docs\u002FWINDOWS.md)\n- [Wrappers and Shell Integration](docs\u002FWRAPPERS.md)\n- [Refactoring Notes](docs\u002FREFACTORING.md)\n\n---\n\n## Detailed Installation\n\n### Setup\n\nIf you want another agent to set up jcode for you, give it this prompt:\n\n```text\nSet up jcode on this machine for me.\n\n1. Detect the operating system, available package managers, and shell environment, then install jcode using the best matching command below instead of referring me somewhere else:\n\n   - macOS with Homebrew available:\n     brew tap 1jehuang\u002Fjcode\n     brew install jcode\n\n   - macOS or Linux via install script:\n     curl -fsSL https:\u002F\u002Fraw.githubusercontent.com\u002F1jehuang\u002Fjcode\u002Fmaster\u002Fscripts\u002Finstall.sh | bash\n\n   - Windows PowerShell:\n     irm https:\u002F\u002Fraw.githubusercontent.com\u002F1jehuang\u002Fjcode\u002Fmaster\u002Fscripts\u002Finstall.ps1 | iex\n\n   - From source if the above paths are not appropriate:\n     git clone https:\u002F\u002Fgithub.com\u002F1jehuang\u002Fjcode.git\n     cd jcode\n     cargo build --release\n     scripts\u002Finstall_release.sh\n\n   - For local self-dev \u002F refactor work on Linux x86_64, prefer:\n     scripts\u002Fdev_cargo.sh build --release -p jcode --bin jcode\n     scripts\u002Fdev_cargo.sh --print-setup\n     scripts\u002Finstall_release.sh\n\n2. Verify that `jcode` is on my `PATH`.\n3. Launch `jcode` once in a new terminal window\u002Fsession to confirm it starts successfully.\n4. Before attempting any interactive login flow, assess which providers are already available non-interactively and prefer those first. Check existing local credentials, config files, CLI sessions, and environment variables such as:\n   - Claude: `~\u002F.jcode\u002Fauth.json`, `~\u002F.claude\u002F.credentials.json`, `~\u002F.local\u002Fshare\u002Fopencode\u002Fauth.json`, `ANTHROPIC_API_KEY`\n   - OpenAI: `~\u002F.jcode\u002Fopenai-auth.json`, `~\u002F.codex\u002Fauth.json`, `OPENAI_API_KEY`\n   - Gemini: `~\u002F.jcode\u002Fgemini_oauth.json`, `~\u002F.gemini\u002Foauth_creds.json`\n   - GitHub Copilot: existing auth under `~\u002F.config\u002Fgithub-copilot\u002F`\n   - Azure OpenAI: `~\u002F.config\u002Fjcode\u002Fazure-openai.env`, `AZURE_OPENAI_*`, or an existing `az login`\n   - OpenRouter: `OPENROUTER_API_KEY`\n   - Fireworks: `~\u002F.config\u002Fjcode\u002Ffireworks.env`, `FIREWORKS_API_KEY`\n   - MiniMax: `~\u002F.config\u002Fjcode\u002Fminimax.env`, `MINIMAX_API_KEY`\n   - Alibaba Cloud Coding Plan: existing jcode config\u002Fenv if present\n5. Prefer whichever provider is already configured and verify it with `jcode auth-test --all-configured` or a provider-specific auth test when appropriate.\n6. Only if no usable provider is already configured, guide me through the minimal manual step needed:\n   - Claude: `jcode login --provider claude`\n   - GitHub Copilot: `jcode login --provider copilot`\n   - OpenAI: `jcode login --provider openai`\n   - Gemini: `jcode login --provider gemini`\n   - Azure OpenAI: `jcode login --provider azure`\n   - Fireworks: `jcode login --provider fireworks`\n   - MiniMax: `jcode login --provider minimax`\n   - Alibaba Cloud Coding Plan: `jcode login --provider alibaba-coding-plan`\n   - OpenRouter: help me set `OPENROUTER_API_KEY`\n   - Anthropic direct API: help me set `ANTHROPIC_API_KEY`\n7. After setup, run a simple smoke test with `jcode run \"say hello\"` and confirm it works.\n8. If I want browser automation, also check `jcode browser status`. If browser automation is not ready, run `jcode browser setup`, verify the built-in `browser` tool works, and explain any remaining manual step.\n9. Explain any manual step that still needs me, especially browser OAuth, device login, API key entry, or browser extension approval.\n```\n\nThis is intended to be a copy-paste bootstrap prompt for jcode itself or any other coding agent.\n\n### Quick Install\n\n```bash\n# macOS & Linux\ncurl -fsSL https:\u002F\u002Fraw.githubusercontent.com\u002F1jehuang\u002Fjcode\u002Fmaster\u002Fscripts\u002Finstall.sh | bash\n```\n\n```powershell\n# Windows (PowerShell)\nirm https:\u002F\u002Fraw.githubusercontent.com\u002F1jehuang\u002Fjcode\u002Fmaster\u002Fscripts\u002Finstall.ps1 | iex\n```\n\n### macOS via Homebrew\n\n```bash\nbrew tap 1jehuang\u002Fjcode\nbrew install jcode\n```\n\n### From Source (all platforms)\n\n```bash\ngit clone https:\u002F\u002Fgithub.com\u002F1jehuang\u002Fjcode.git\ncd jcode\ncargo build --release\n```\n\nFor local self-dev \u002F refactor work on Linux x86_64, prefer:\n\n```bash\nscripts\u002Fdev_cargo.sh build --release -p jcode --bin jcode\nscripts\u002Fdev_cargo.sh --print-setup\n```\n\nThat wrapper automatically uses `sccache` when available, prefers a fast\nworking local linker setup (`clang + lld`) instead of assuming every machine's\n`mold` configuration is valid, and can print the active linker\u002Fcache setup via\n`--print-setup` so slow-path builds are easier to diagnose.\n\nThen symlink to your PATH:\n\n```bash\nscripts\u002Finstall_release.sh\n```\n\n### Platform Support\n\n| Platform | Status |\n|---|---|\n| **Linux** x86_64 \u002F aarch64 | Fully supported |\n| **macOS** Apple Silicon & Intel | Supported |\n| **Windows** x86_64 | Supported (native + WSL2) |\n\n\u003C\u002Fdiv>\n","jcode 是一个用于提升编程技能上限的下一代编码代理工具。它支持多会话工作流，具有高度可定制性和出色的性能表现。项目采用 Rust 语言开发，注重资源效率和启动速度，在处理单个活跃会话时，其内存占用仅为27.8 MB（本地嵌入关闭情况下），显著优于同类工具。jcode 适用于需要高效执行复杂或大规模代码操作的开发者，特别是在对系统资源有严格要求的环境中。",2,"2026-06-11 02:31:42","trending"]