# Claw-Empire: An AI Agent Office Simulator for Orchestrating Multiple Coding Assistants

> Claw-Empire is a local-first TypeScript application that wraps Claude Code, Codex CLI, Gemini CLI, OpenCode, and other AI coding agents in a virtual office metaphor, giving each agent a department role, a task queue, and a pixel-art dashboard. The last push to the repository was on 2026-03-16.

**GreenSheep01201/claw-empire** — Command Your AI Agent Empire from the CEO Desk — A local-first AI agent office simulator that orchestrates CLI, OAuth, and API-connected agents (Claude Code, Codex CLI, Gemini CLI, OpenCode, and more) as a virtual autonomous company.

- Repository: https://github.com/GreenSheep01201/claw-empire
- Stars: 1,389 · Forks: 267
- Language: TypeScript
- License: Apache-2.0
- Published: 2026-09-10 · Updated: 2026-09-10 · Language: en
- Canonical page: https://hysenlabs.com/projects/greensheep01201-claw-empire

## What Claw-Empire Does and Why the Office Metaphor

Claw-Empire presents itself as a virtual software company where the developer is the CEO and the AI coding assistants are employees. The metaphor is not purely decorative: each agent is assigned to a department, receives tasks through a kanban board, reports completions through a messenger integration, and accumulates skills from a library of over 600 skill definitions.

The practical value is having a single interface for managing multiple AI agents that would otherwise require opening and monitoring separate terminals. Claude Code, Codex CLI, Gemini CLI, OpenCode, Kimi Code, GitHub Copilot, and Antigravity are all listed as supported providers. Agents connect through three distinct methods: CLI execution, OAuth authentication, and direct API key. All application data stays local in a SQLite database with no cloud dependency.

The pixel-art office visualization is the most visually distinctive feature. It is not just a skin over a task manager; the README describes it as a way to make AI orchestration transparent and to show which agents are active, idle, or completing tasks. Whether this framing improves productivity depends on how you think about multi-agent workflows, but the underlying kanban, skill assignment, and messenger integration features work independently of the visual layer.

## Three Connection Methods for Different Provider Types

The provider connection model is one of the more practical design decisions in Claw-Empire. Different AI coding tools expose themselves differently: some require a CLI binary on PATH, some authenticate through OAuth, and some accept an API key. Claw-Empire accommodates all three without forcing every provider through the same abstraction.

CLI providers like Claude Code and Gemini CLI run as child processes. The Dockerfile installs the required binaries at build time:

```bash
npm install -g \
  @anthropic-ai/claude-code \
  @openai/codex \
  @google/gemini-cli \
  opencode-ai
```

OAuth providers authenticate through the settings UI using GitHub or Google OAuth. The application encrypts and stores the resulting tokens in SQLite using the OAUTH_ENCRYPTION_SECRET environment variable. This key must be set before any OAuth provider can be configured. The .env.example generates the required value:

```bash
node -e "console.log(require('crypto').randomBytes(32).toString('hex'))"
```

Direct API providers like OpenCode Go accept a base URL and API key through the settings panel. The v2.0.4 release notes describe new presets for direct API providers that lock the expected base URL and seed default models, reducing the manual configuration required for each new agent.

## Office Packs: Six Collaboration Topologies

Claw-Empire ships six office pack profiles that configure the department structure, naming scheme, and workflow bias for different types of work. The development pack is the default engineering baseline, with departments for planning, development, design, QA, DevSecOps, and operations.

The report pack configures the office for structured document production, with an editorial planning department, a research engine, a document design team, and a review desk. The web research report pack adds a fact-checking step and source collection. The novel pack configures worldbuilding, narrative, character art, and tone QA departments for long-form writing. The video preproduction pack covers concept, script, shot list, and editing notes. The roleplay pack enables in-character dialogue with consistency checking.

Each pack applies a different naming seed and workflow bias, but all packs share the same kanban and skill assignment infrastructure. Switching packs is a configuration change rather than a reinstall. The development pack expands to initialized office packs as the default baseline when assigning models to agents, according to the v2.0.4 release notes.

## Getting Started: Docker Deployment

The recommended deployment path for Claw-Empire is Docker, which the README describes as an officially documented path as of v2.0.4. The docker-compose.yml is in the repository root and exposes the application on port 8790:

```bash
docker compose up -d
```

Before starting the container, create a .env.docker and .env.docker.private file with at least the required variables. The .env.example documents all available variables. The minimum required variable is OAUTH_ENCRYPTION_SECRET, which encrypts OAuth tokens stored in SQLite. Without it, OAuth provider authentication will fail.

The Docker image installs the CLI provider binaries at build time, which means Claude Code, Codex CLI, Gemini CLI, and opencode-ai are available inside the container without additional setup. The container runs as an unprivileged user (UID 10001) and mounts the data/ directory for the SQLite database and logs. The security_opt line in docker-compose.yml sets no-new-privileges:true.

For developers who want to run without Docker, the native dev server starts with pnpm dev, which launches both the Vite frontend on port 8800 and the Node.js API server with nodemon watching the server/ directory.

## Messenger Integration and the Dollar Command

One of the more unusual features in Claw-Empire is the messenger integration. The application can send and receive messages through Telegram, WhatsApp, Discord, Google Chat, Slack, Signal, and iMessage. The CEO can send directives to agents using the $ command prefix from any of these channels.

This is configured through the settings UI rather than environment variables, and messenger tokens are stored encrypted in the SQLite database alongside OAuth tokens. The INBOX_WEBHOOK_SECRET variable in .env.example enables the /api/inbox webhook endpoint that receives incoming messages from messenger services.

The practical use case is asynchronous task delegation: you send a $ directive from your phone, the agent processes it, and the completion report arrives back in the same messenger thread. The meeting minutes feature generates AI summaries of multi-round discussions with approval steps before the summary is finalized.

## Maintenance State and What to Verify Before Adopting

The last push to the repository was on 2026-03-16, and the most recent release is v2.0.4 from 2026-03-12. That is six months and twelve days before this article was written. The repository is not currently receiving updates.

For a project at v2.0.4 with a meaningful feature set, a six-month gap may reflect maintenance stability rather than abandonment, but it may also reflect a project that has stalled. The SECURITY.md and the CONTRIBUTING.md files are present, and the release notes are detailed, suggesting active development at the time of the last release. Whether development resumes is unknown.

The license is Apache-2.0. The application stores all data in a local SQLite file at the path specified by DB_PATH, defaulting to ./claw-empire.sqlite. There is no telemetry or external service dependency in the core application; the only external connections are to the provider APIs you configure. Before exposing Claw-Empire to a non-loopback address, review SECURITY.md and set API_AUTH_TOKEN to protect the API from unauthorized access, as the .env.example notes that remote browser clients provide the token at runtime and it is not embedded in build artifacts. Log files are written to the directory set by LOGS_DIR, which also defaults to a local path.

## Conclusion

Claw-Empire is a reasonable choice for developers who work with multiple AI coding agents simultaneously and want a single dashboard to track tasks, assign skills, and review output. The office and CEO metaphor is intentional rather than cosmetic: it provides a shared language for describing what each agent is doing. The last push was on 2026-03-16, which is six months and twelve days before this article was written, so the project is not currently receiving updates. Developers who need active maintenance should verify whether the repository resumes activity before building workflows around it. The Docker deployment path and the SQLite database make it straightforward to run a self-contained instance; check the SECURITY.md before exposing the API to non-loopback addresses.

## FAQ

### Who is Claw AI?

The README describes Claw-Empire as a project for orchestrating AI coding agents from a single dashboard. The repository is authored by GreenSheep01201 on GitHub. The README does not document a company or organization named Claw AI behind the project.

### Does Claw-Empire support running agents in isolated git worktrees?

Yes. The README describes worktree terminals that run under .octogent/worktrees/ style isolation. Agents working in worktree mode operate on octogent-prefixed branches, keeping their changes separate from the main workspace until they are reviewed.

### What database does Claw-Empire use and does it send data to the cloud?

Claw-Empire uses a local SQLite database stored in the path set by the DB_PATH environment variable. All application data stays on the local machine. The application is described as local-first with no cloud dependency for core functionality.

## Sources

- [GreenSheep01201/claw-empire on GitHub](https://github.com/GreenSheep01201/claw-empire)
- [Issues](https://github.com/GreenSheep01201/claw-empire/issues)
- [License: Apache-2.0](https://github.com/GreenSheep01201/claw-empire/blob/main/LICENSE)
- [README](https://github.com/GreenSheep01201/claw-empire/blob/main/README.md)
- [Releases](https://github.com/GreenSheep01201/claw-empire/releases)

---

Hysen Labs editorial analysis, written from the project's own repository and release notes. Cite the canonical page: https://hysenlabs.com/projects/greensheep01201-claw-empire