humanlayer/humanlayer

The best way to get AI coding agents to solve hard problems in complex codebases.

HumanLayer is a system that runs AI coding agents on your own machines and gives your team a single place to steer them. The agents (such as Claude Code, Codex, OpenCode, or Pi) execute on your computers, while tasks, sessions, plans, and reviews stay in sync between a desktop app, a web app, and a background program (a “daemon”) on each machine. A team can watch and direct agent work from one unified interface instead of juggling separate terminals.

What the repo contains

This is a monorepo — one repository holding many related apps and libraries — built with Turborepo and the Bun runtime. The pieces:

  • User-facing apps — the main UI (available both as a web app and inside desktop shells), plus macOS desktop apps written with Electron and Tauri.
  • The daemon — apps/riptide-daemon runs agent sessions on a host machine. It is also shipped inside the humanlayer command-line tool (apps/cli).
  • Backend pieces — an API built with oRPC and Hono, and an authentication proxy in front of Electric (a tool that syncs Postgres data out to client apps).
  • Admin web app — apps/riptide-cloud handles organizations, invites, settings, and the install page.
  • Housekeeping jobs — several AWS Lambda functions: a database migrator, an updater server, a stale-daemon sweeper, and a database-maintenance function.
  • Plugins & extensions — plugins that mirror Claude Code, Codex, OpenCode, and Pi sessions back into HumanLayer, plus the “Research-Plan-Implement” (RPI) skills with their own CLIs, and a Linear CLI.
  • Docs and dev tooling — user documentation, and apps/devstack, which runs a complete local stack for development.
  • Shared code — packages/ holds libraries (database, schemas, UI, sync, AI, billing, etc.), and contracts/ holds the RPC contracts that define how the apps talk to each other.

Getting started

To build from source you need Bun and Docker (and Rust, if you build the Tauri desktop app). The basic steps:

bun install
cp .env.example .env
bun --bun run typecheck
bun run test

bun --bun run devstack help explains how to run the whole stack locally. CONTRIBUTING.md lists every check a pull request must pass.

Privacy note

Building from source sends nothing back to the project maintainers. Tools like Sentry (error tracking), PostHog (analytics), and the WorkOS client ID are all read from environment variables, so they simply stay off unless you set those variables yourself. The .env.example file lists them.

Self-hosting and license

HumanLayer Cloud runs this same code. Official guides for self-hosting are still being written. The project is licensed under Apache License 2.0, with a few vendored folders keeping their own MIT licenses (listed in the NOTICE file).

Related

  • Project
    zilliztech/claude-context

    An MCP plugin that adds semantic code search to AI coding agents, allowing them to efficiently retrieve relevant snippets from large codebases using a vector database.

  • Project
    openchamber/openchamber

    OpenChamber is an open‑source, cross‑platform workspace that lets developers run AI coding agents, set automated goals, compare multiple model runs, review changes with guided walkthroughs, and ship code—all from desktop, web, VS Code, or mobile.

  • Project
    The-Pocket/PocketFlow-Tutorial-Codebase-Knowledge

    An AI-powered tool that analyzes GitHub repositories or local codebases to automatically generate beginner-friendly tutorials and visualizations explaining how the code works.

  • Project
    Piebald-AI/claude-code-system-prompts

    A repository that extracts, version‑tracks, and publishes every system‑prompt string used by Anthropic’s Claude Code code‑assistant, providing raw prompt files, token counts, and a changelog for over 500 fragments.