Skip to content

patleeman/neon-pilot

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

7,573 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Neon Pilot

Neon Pilot logo

A self-extensible desktop agent harness, powered by the Pi Coding Agent.

Website · Docs · Download for macOS


Neon Pilot is a durable AI agent runtime with a desktop app, background automation, extensions, and knowledge management. It is built for agents that can improve their own workspace instead of waiting for the next product release.

Ask for a workflow once, then keep it: commands, tools, automations, UI surfaces, skills, and local extensions are first-class runtime capabilities.

Start here

What it does

  • Self-extensible workflows — ask Neon Pilot to add commands, tools, pages, panels, settings, or reusable skills.
  • Extension-first product surface — product workflows live in system or user extensions instead of bloating core.
  • Durable conversations — conversations, transcripts, branches, checkpoints, artifacts, and background work persist across app restarts.
  • Background work — daemon-backed background commands, subagents, automations, scheduled tasks, follow-ups, and deferred resumes.
  • Knowledge base — git-backed durable knowledge, instruction files, skills, project notes, and managed sync.
  • Multi-provider models — bring your own providers and pick the right model for each task.
  • Workbench UI — conversation view plus workbench panes for files, diffs, artifacts, browser surfaces, knowledge, and extension tools.

Installation

Download the latest macOS .dmg from GitHub Releases, open it, and drag Neon Pilot.app into Applications.

For agent-driven install and bootstrap on macOS, use the packaged installer flow:

curl -fsSL https://raw.githubusercontent.com/patleeman/neon-pilot/master/install.sh | bash -s -- --install-cli --bootstrap --json

Then configure and verify through the CLI:

neon-pilot bootstrap configure --secrets-provider keychain --provider openai-codex --model gpt-5.4 --json
printf '%s' "$OPENAI_API_KEY" | neon-pilot bootstrap provider set-key openai --stdin --json
neon-pilot bootstrap doctor --json

Open Neon Pilot.app, start a new conversation, and send a small prompt such as “Summarize this app setup.” The first-run path is complete when the app loads, doctor passes, and the agent replies.

See Getting Started for the user setup flow and first task checklist. See Agent bootstrap for the complete external-agent setup contract, including provider setup, verification commands, and Hermes/MCP configuration.

Development

pnpm install
pnpm run setup:hooks   # optional: enable tracked git hooks
pnpm run build
pnpm run desktop:start

For the desktop dev app:

pnpm run desktop:dev

For the website:

pnpm run build:site
cd apps/site && python3 -m http.server 4173

Architecture

Neon Pilot keeps core small and pushes user-facing behavior into extensions.

  • packages/core — shared runtime primitives, resource resolution, prompt assembly inputs, conversation/storage utilities.
  • packages/desktop — Electron app shell, desktop/server runtime, local API, extension host, daemon integration, UI.
  • packages/extensions — public extension SDK types and frontend/backend seams.
  • extensions/system-* — bundled first-party product features.
  • patleeman/neon-pilot-extensions — optional first-party extensions that install into runtime state as user extensions.
  • apps/site — static marketing/docs site published at neonpilot.net.
  • docs — agent-facing product, architecture, extension, and development docs.

Documentation

Start with neonpilot.net/docs or docs/README.md.

Important repo docs:

License

MIT