Autonomous AI Teams
That Build Software
Deploy a full team of AI agents — product owner, architect, engineers, QA, SRE — that collaborate to spec, design, implement, review, test, and deploy your software. Fully containerized. Self-hosted and free.
Plug and Play
Clone, add an API key, docker compose up. No Kubernetes, no cloud accounts, no hour-long setup. Your AI team is running in under 5 minutes.
Real Team, Not a Chatbot
Each agent has a rich persona with backstory, opinions, and domain expertise. Eric pushes back on vague specs. Finn rejects bad architecture. Chieko finds the edge cases you forgot.
Fully Containerized
Every agent runs in its own isolated Docker container with a full engineering toolbox — Node, Python, Go, Rust, git, and dozens more. No host access, no security concerns.
Persistent Memory
Agents remember decisions, lessons, and patterns across runs using ClawVault with semantic search. They learn from mistakes and improve over time.
Beautiful Dashboard
Real-time streaming output, project management with kanban boards, chat interface, pipeline visualization, and agent configuration — not a terminal dump.
Slack-Native
Each agent gets its own Slack bot. Watch your team discuss in threads. Or skip Slack entirely and use the built-in chat interface.
YAML Pipelines
Define workflows as simple YAML — steps, agents, branching, loops, retries. No code required for orchestration. Create custom pipelines for any workflow.
MCP Tools
Agents use any MCP-compatible tool server via the built-in mcpo proxy. Add GitHub, web search, or any custom tool with a single config entry.
Open Core
Self-hosted is completely free. Use it, modify it, run it on your own infrastructure. FSL-1.1-ALv2 license converts to Apache 2.0 after 2 years.
1. Create a project — describe what you want built via the dashboard’s guided onboarding, or import an existing repo.
2. Plan it — the planning pipeline decomposes your project into tasks on a kanban board with priorities and dependencies.
3. Assign agents to columns — Yukihiro watches “Ready” for implementation work, Chieko watches “Review” for testing, Stas watches for deployment. Each agent knows their lane.
4. Agents work autonomously — on pulse cycles, agents wake up, check the board, claim a task, spin up an isolated container, do the work, open a PR, and advance the task.
5. Watch it happen — streaming output in the dashboard, Slack threads, or the kanban board. Step in when you want, or let them run.
Pipelines handle structured workflows — the planning pipeline generates the task board, the engineering pipeline runs a full SDLC for a single task, and you can build custom pipelines for any repeatable process.
| Agent | Role | Pipeline Stage | What They Do |
|---|---|---|---|
| Eric | Product Owner | SPEC | Requirements, user stories, acceptance criteria, scope management |
| Finn | Solutions Architect | DESIGN / REVIEW | System architecture, tech decisions, code review, API design |
| Shigeo | UX Specialist | UX | User flows, design systems, component specs, accessibility |
| Yukihiro | Senior SWE | IMPLEMENT / FIX | Implementation, bug fixes, writing production code |
| Chieko | Test Engineer | TEST | QA strategy, regression detection, test automation |
| Stas | SRE | DEPLOY | Infrastructure, deployment, monitoring, incident response |
| Yang | DevEx Specialist | DX (on-demand) | CI/CD pipelines, tooling, developer workflow optimization |
| Holt | Marketing & Sales | On-demand | Sales strategy, outreach, deal management, positioning |
| Luke | SEO Specialist | On-demand | Content strategy, keyword research, technical SEO |
| Jim | Finance Lead | On-demand | Budget, pricing, runway management, financial modeling |
The engineering pipeline is fully operational today. Marketing, sales, and finance agents work in chat and pulse modes, with structured pipeline support coming soon.
| DjinnBot | OpenClaw | Typical Agent Frameworks | |
|---|---|---|---|
| Setup time | docker compose up — 5 minutes | Kubernetes + cloud config — hours | Framework wiring + custom code — hours to days |
| Interface | Full dashboard, Slack bots, chat, CLI, API | Basic web UI | Terminal output or minimal web UI |
| Security | Every agent in isolated Docker container. No host access. | Direct host access, shell execution | Direct host access, shell execution |
| Agent memory | Persistent semantic memory with knowledge graph across runs | Stateless or basic file storage | Stateless or basic file storage |
| Multi-agent collaboration | Agents review, critique, and build on each other's work | Loose coordination | Single-agent or sequential handoff |
| Customization | YAML pipelines, markdown personas — no code | Code-level changes | Code-level changes |
| Agent personas | Rich characters with opinions, beliefs, and anti-patterns | Generic system prompts | Generic system prompts |
| Tool system | MCP tools converted to native tools at runtime | Manual tool configuration | Custom tool integrations |
DjinnBot is built for people who want autonomous AI teams working on real projects — not another framework to wire together.