Self-hosted AI coding control plane

Run every coding agent from one place.

Pixcode turns your machine or server into a live workspace for AI coding CLIs. It keeps chat, files, shell, Git/local changes, orchestration, TaskMaster planning, Telegram, notifications, MCP, plugins, and API automation in one browser or desktop app.

npx @pixelbyte-software/pixcode
open http://localhost:3001
Why it exists

Coding agents are powerful, but terminals alone do not show enough.

Pixcode adds the missing operating layer around provider-native CLIs: sessions, visible progress, file changes, shell output, source control, orchestration history, task planning, remote control, and API access.

What Pixcode can do

It is not a landing page wrapper around one chat box. It is a practical workspace for real projects, long-running agent work, and multiple providers.

01

Talk to coding CLIs

Use Claude Code, Cursor CLI, Codex, Gemini CLI, Qwen Code, and OpenCode from the same project screen.

02

See what changed

Track Git changes or local filesystem changes, highlight edited files, and open the changed location from the side rail.

03

Run shell and inspect files

Keep the main chat visible while opening files, shell, source control, and project tools in responsive split panels.

04

Orchestrate teams

Run agent teams, sequential handoffs, multi-model reviews, and decision debates with per-agent roles and models.

05

Plan with TaskMaster

Enable PRD parsing, task planning, task status, and agent task execution with provider keys or a custom OpenAI-compatible API.

06

Automate externally

Use REST, WebSocket, and `px_` API keys to connect scripts, services, dashboards, CI, or Telegram to the same backend.

07

Work remotely

Run Pixcode on a workstation or server, keep sessions alive, and control them from a browser, desktop app, or Telegram.

08

Customize the tool

Use themes, MCP management, plugin surfaces, provider settings, notifications, and model controls without rebuilding the app.

Agent orchestration

When one agent is not enough, make the workflow explicit.

Pixcode can split work across roles, pass compact context between stages, compare outputs, and preserve the run history so you can understand what each agent did and why.

Agent TeamParallel or staged workers for implementation, review, docs, and custom roles.
Sequential HandoffEach step receives a compact summary of the previous step instead of a blind prompt.
Multi-model ReviewCompare provider and model opinions before accepting an implementation.
Fallback AgentChoose a backup CLI for failed orchestration steps.
TaskMaster and custom APIs

Use known providers or bring any OpenAI-compatible endpoint.

TaskMaster settings can store standard provider variables and custom OpenAI-compatible API URL, API key, and model values. That lets a private gateway, local router, hosted model provider, or company endpoint power task planning without hardcoding one provider.

OPENAI_API_KEY=sk-custom-...
OPENAI_BASE_URL=https://your-provider.example.com/v1
OPENAI_MODEL=your-model-id

Built to be useful as an open-source project

The repository includes the project files people expect before they trust, install, contribute to, or recommend a developer tool.

README with clear purpose and install commands Open-source LICENSE CONTRIBUTING guide CODE_OF_CONDUCT SECURITY policy Issue templates and good-first-issue template Screenshots and static docs Releases, tags, npm package, and desktop assets OpenAPI, llms.txt, sitemap, robots.txt, and SEO pages

Install it where the work happens

Use npm for a server/local install, or download desktop builds from GitHub Releases. Normal installed usage serves the frontend and backend from port 3001.

npx @pixelbyte-software/pixcode
npm install -g @pixelbyte-software/pixcode
pixcode
pixcode daemon install --mode auto --port 3001
pixcode daemon status --mode auto