---

px0 is an IDE built for reviewing AI-generated code, optimized for speed. It turns your browser into a zero-latency console with native Git and GitHub integrations, instant search across massive codebases, and seamless handoff to local AI coding harnesses.

As AI agents author more code directly from the terminal, engineering productivity is no longer constrained by how fast you type - it is constrained by how fast you can review, navigate, and verify changes. px0 replaces heavy, multi-gigabyte editing suites with an instant, distraction-free environment built specifically for this review loop.

• < 1 ms cold start
• ~20–30 MB server daemon RAM (~100–180 MB total with browser tab, vs. VS Code's ~1,440 MB)
• Single static Go binary with zero runtime dependencies (no Electron, no Node, no CGO)
• ~280 languages tokenized natively via Chroma

See full performance benchmarks and comparisons at px0.ai/benchmarks.

Features

• GitHub PR reviews & Git panel: Review pull requests directly (px0 <pr-url>), inspect scoped merge-base diffs, draft inline review comments, and stage, commit, or push from the browser.
• Unpushed commit review: Read the commits your branch is ahead of its tracking branch by, one file diff at a time, before you push them.
• AI coding harness integration: Dispatch edits directly to Claude Code, Gemini CLI, Cursor Agent, Antigravity, OpenCode, Codex, Aider, or Goose with live reloading.
• Fast navigation: Fuzzy file search, symbol outline, and workspace regex search in milliseconds.
• Remote-first: Run on any remote server, VM, or container and browse locally without SSH key setups or remote daemons.
• Virtual rendering: Opens 400,000-line files smoothly by mounting only visible rows; frees memory back to the OS after 15 seconds of inactivity.
• Rich code viewer: 14 built-in themes, rendered Markdown preview, CSV/TSV table view, image inspector, and optional zero-config LSP for Go-to-Definition and hover.

Installation

Quick install (macOS, Linux, BSD):

curl -fsSL https://px0.ai/install.sh | sh

Build from source (requires Go 1.24+ and Node.js or Bun):

git clone https://github.com/px0-ai/px0.git
cd px0
make build
install -d ~/.local/bin && install px0 ~/.local/bin/

To update an existing installation:

px0 --update

Usage

# Inspect current directory
px0

# Inspect a specific directory or file
px0 ~/workspace/project
px0 main.go:42

# Review a GitHub pull request
px0 https://github.com/owner/repo/pull/123

# Remote or headless server mode
px0 -host 0.0.0.0 -port 7777 ~/workspace

# Behind a reverse proxy under a subpath
px0 -base-path /rev-123/ -host 0.0.0.0 -port 7777 ~/workspace

When px0 binds to 0.0.0.0, it prints a network URL for every unique non-loopback IPv4 address on the machine, using the port selected by the listener:

  url:        http://0.0.0.0:7777
  network:    http://10.0.0.15:7777
  network:    http://192.168.1.42:7777

Open the address that is reachable from your local machine. The list can include LAN, VPN, and container-network addresses, depending on the remote host's interfaces.

Access securely over Tailscale, WireGuard, reverse proxy, or Cloudflare Tunnel with zero remote setup overhead and strict read-only sandboxing (path traversal protection & DNS rebinding checks).

Development

git clone https://github.com/px0-ai/px0.git
cd px0

# Run tests
make test

# Live frontend development (serves web/ assets directly from disk)
go run . -dev .

# Build binary
make build

Benchmarks

# Clone the benchmark corpus (~3 GB: Linux, Kubernetes, TypeScript, etc.)
./benchmark.sh --clone

# Run benchmarks across the corpus
./benchmark.sh

# Compare px0 directly against VS Code on the current workspace
./benchmark.sh --vscode .

# Profile memory lifecycle (index, search, idle memory release)
./benchmark.sh --memory bench-repos/linux

Architecture & Memory Accounting: px0 separates the lightweight native Go daemon from the browser-rendered UI. The Go server consumes ~20–30 MB resident RAM (RSS). The active browser tab allocates ~80–150 MB for DOM nodes, V8 JS runtime, and GPU compositing. Combined, the total local memory footprint is ~100–180 MB (~85–90% lighter than Electron IDEs like VS Code at ~1,440 MB, which bundle dedicated Chromium and Node runtimes). In remote or container environments (px0 -host 0.0.0.0), the remote host pays strictly the ~20–30 MB server cost. See BENCHMARKS.md for full methodology and breakdowns.

Documentation

• Docs
• Benchmarks
• Blog
• Changelog
• Partners

Design Partners

When AI coding agents generate large diffs across repositories every day, code authoring is no longer the bottleneck. Verification and review are.

We are looking for engineering teams (10+ engineers or teams running active agent workflows) as design partners. Partners receive direct Slack Connect access to the core team, rapid turnaround on custom harness integrations, and direct input on the roadmap.

Learn more on the Design Partners page.

License

MIT License © Arpit Bhayani