Human-free software development with self-organizing AI agent teams.
- Human-free execution — Agents plan, discuss, research, and implement autonomously across full development cycles
- Self-organizing teams — AI managers (Athena, Ares, Apollo) plan, implement, and verify milestones without human intervention
- Multi-project — Manage multiple repos from one central orchestrator with independent cycles
- Multi-provider — Anthropic, OpenAI, Google, Groq, Mistral, xAI, Amazon Bedrock, Azure OpenAI, Cerebras, HuggingFace, MiniMax, OpenRouter, GitHub Copilot, Kimi Coding, and custom endpoints
- Key pool — Multiple API keys with priority ordering, automatic cross-provider fallback, and rate limit cooldowns
- Full observability — Every decision, discussion, and code change is visible through the dashboard and GitHub PRs
- Chat — Talk to agents directly through the dashboard with streaming responses and image support
- Async human intervention — Agents escalate via issues when they need human input; step in at your convenience
- Budget controls — 24-hour rolling budget limiter with per-agent cost tracking and computed sleep intervals
- Unified dashboard — Monitor all projects, agents, issues, and PRs in one place (mobile-friendly, dark mode, push notifications)
Monitor agents, costs, issues, and reports across all your projects from a single dashboard.
- Node.js ≥ 20
- GitHub CLI (
gh) — installed and authenticated - An API key from any supported provider (Anthropic, OpenAI, Google, etc.)
# Install globally
npm install -g thebotcompany
# Start the orchestrator + dashboard (first run will prompt for password and port)
tbc startAdd your API key in Settings, create a project through the dashboard UI, then start the orchestrator.
tbc start # Start orchestrator + dashboard (background)
tbc stop # Stop orchestrator
tbc dev # Start in dev mode (foreground + Vite HMR)
tbc status # Show running status
tbc logs [n] # Show last n lines of logs (default 50)The dashboard provides:
- Project overview — Status, phase, milestone progress, epoch/cycle count
- Agent reports — Full history of agent outputs with markdown rendering and auto-summarization
- Agent issues — SQLite-backed issue tracker for agent-to-agent coordination
- Human intervention — Separate panel for human↔agent escalation issues
- Chat — Direct conversation with agents per project, with streaming and image uploads
- PR monitoring — Live GitHub PR status per project
- Cost tracking — Per-agent and per-project cost breakdown (last cycle, average, 24h, total)
- Controls — Pause/resume, skip sleep, kill run/cycle/epoch, bootstrap
- Key pool management — Add/remove/reorder API keys, OAuth sign-in, per-project key selection with fallback control
- Model tier overrides — Customize which model runs at each tier (high/mid/low/xlow) per project
- Notifications — Browser push notifications for milestones, verifications, and errors
- Settings — Theme (light/dark/system), credential management, notification preferences
The dashboard has read-only mode by default. Enter the password (set during first-run setup) via the login button to enable write operations (pause, resume, config changes, etc.).
| Variable | Default | Description |
|---|---|---|
TBC_PASSWORD |
(set on first run) | Password for write access to the dashboard |
TBC_PORT |
3100 |
Server port |
TBC_HOME |
~/.thebotcompany |
Data directory (keys, DB, logs) |
TBC_SERVE_STATIC |
true |
Serve the built dashboard frontend |
TBC_ALLOW_CUSTOM_PROVIDER |
false |
Enable custom provider support (see Security) |
API keys can be added through the dashboard Settings panel. Alternatively, set environment variables for auto-detection on first run:
ANTHROPIC_API_KEY=sk-ant-...
OPENAI_API_KEY=sk-...
GEMINI_API_KEY=...The key pool supports multiple keys per provider with automatic fallback when a key hits rate limits.
| Provider | Auth Methods |
|---|---|
| Anthropic | API key, OAuth |
| OpenAI | API key, OAuth |
| Google (Gemini) | API key, OAuth |
| GitHub Copilot | OAuth |
| Amazon Bedrock | API key |
| Azure OpenAI | API key |
| Cerebras | API key |
| Google Vertex | API key |
| Groq | API key |
| Hugging Face | API key |
| Kimi Coding | API key |
| MiniMax | API key |
| Mistral | API key |
| OpenRouter | API key |
| xAI | API key |
| Custom | API key (OpenAI or Anthropic compatible) |
- Password-protected writes — All mutating API endpoints require authentication via
TBC_PASSWORD. Read-only access (viewing projects, logs, reports) does not require auth. - CORS — API only accepts requests from the dashboard origin (
localhost:<port>).
The custom provider feature allows connecting to any OpenAI-compatible or Anthropic-compatible API endpoint. This introduces a server-side request forwarding (SSRF) surface since TBC's server makes HTTP requests to the user-specified baseUrl.
Built-in protections:
- Private IP blocklist —
baseUrlis validated and rejected if it points to:- Localhost (
127.0.0.1,::1,localhost) - RFC1918 private ranges (
10.0.0.0/8,172.16.0.0/12,192.168.0.0/16) - Link-local addresses (
169.254.0.0/16) - Cloud metadata endpoints (
169.254.169.254,metadata.google.internal)
- Localhost (
- Protocol restriction — Only
http://andhttps://URLs are accepted. - Auth-gated — Creating or editing custom credentials requires write authentication.
The custom provider is disabled by default. To enable it:
# In your .env or environment
TBC_ALLOW_CUSTOM_PROVIDER=trueWhen disabled, custom credential creation is blocked at the API level and the option is hidden from the UI.
Note: The hostname blocklist does not protect against DNS rebinding attacks (where an external hostname resolves to a private IP). For maximum security on shared instances, keep
TBC_ALLOW_CUSTOM_PROVIDERdisabled (the default).
# Clone the repo
git clone https://github.com/syifan/thebotcompany.git
cd thebotcompany
# Install dependencies
npm install
cd monitor && npm install && cd ..
# Run in dev mode (server + Vite HMR)
tbc dev
# Or run components separately
node src/server.js # Server only
cd monitor && npm run dev # Dashboard only (proxies API to :3100)
# Build dashboard for production
cd monitor && npm run build
# Run tests
npm test