Plugin-driven multi-tenant AI agent runtime. Single Go binary with .so plugins, supporting multiple LLM backends and chat platforms.
- Multi-tenant architecture with per-tenant agents, users, and invite codes
- Plugin system for LLM backends, chat channels, tools, sandbox, and middleware
- Docker sandbox for isolated code execution with container hardening
- Built-in tools: web search, shell, memory, scheduling, file access, and more
- Chat platform adapters (Telegram, Microsoft Teams)
- CLI for tenant, agent, user, and workspace management
cp config.example.json data/config.json
touch .env
# Edit data/config.json and .env with your API keys
make dind-certs
docker compose up -d --build| Topic | Description |
|---|---|
| CLI Reference | Tenant, agent, user, invite, and workspace commands |
| Configuration | Config sections and environment variable syntax |
| Docker Workflows | Running in Docker, common recipes |
| Deployment | DinD (default) and bare metal strategies |
| Security Model | Sandbox hardening and deployment tradeoffs |
Prerequisites: Go 1.24+, CGO enabled.
make build # Build binary to bin/whiteagent
make plugins # Build plugin .so files
make test # Run all tests
make fmt # Format code
make clean # Remove bin/Run make help for all targets.
cp config.example.json config.json
cp .env.example .env
make build plugins
./bin/whiteagent serve --config config.json