Sulala Agent OS is a lightweight Bun-based Agent Operating System for:
~/.agent-os/skills/ or a storeDesigned to be modular, local-first, and extensible.
| Phase | Focus | Status |
|---|---|---|
| 1. Core runtime | Bun server, agent registry, LLM loop, single-agent run | Done |
| 2. Tool system | Tool registry, built-in tools (echo, time), tool execution in loop | Done |
| 3. Skill system | Skill loader (skill.yaml / SKILL.md, tools.yaml), install, config | Done |
| 4. Task queue & scheduler | Queue, workers, cron, persist tasks | Done |
| 5. Event system | Event bus, runtime events, WebSocket | Done |
| 6. Agent graph | Graph manager, runner, multi-agent workflows | Done |
| 7. API layer | REST + WebSocket, CORS, validation | Done |
| 8. Dashboard | Agents, Tasks, Skills, Graphs, Settings, Chat | Done |
| 9. Security & limits | Workspace sandbox, runtime limits, allowed tools | Done |
| 10. Conversation & memory | Conversations, summarization, memory skill + API | Done |
Additional shipped features: Channels (Telegram, Slack, Discord, Signal, Viber), memory UI, AI provider selection (OpenRouter, OpenAI, Anthropic, Google, Custom), agent templates on create, default agents (personal, research, writer, etc.), skill marketplace and store registry API.
The source of truth for phases, checklists, and future work is the Agent OS repo:
sulala/roadmap.md — Full roadmap: vision, design principles, tech stack, home directory layout, components (Agent Manager, Runtime, Tool Registry, Skill Loader, Task Queue, Scheduler, Event Bus, Worker Pool, Memory, Graphs), API layer, dashboard, security, example agents, development phases, and future improvements (distributed workers, more skills, etc.).Refer to that file for:
| Area | Doc (in repo) |
|---|---|
| Agent OS roadmap (phases, future) | sulala/roadmap.md |
| Agent API & endpoints | Agent API, Agent OS |
| Skills & store | Skills authoring, Store |
| CLI | CLI reference |