Documentation
Roadmap
6-phase plan from operational foundation to production MVP launch. Current: Phase 1.
Vision
ClawPilot aims to be the transparent, vendor-agnostic operating system for autonomous AI agents. Not a wrapper around one LLM, but a full infrastructure layer: task management, multi-agent routing, skills marketplace, observability, and compliance — all open source and self-hostable.
Phase 1 — Operational Foundation
In Progress — March 2026
- Hub dashboard with real-time SSE stream
- Kanban API (CRUD, dependency enforcement, confidence scoring)
- Append-only audit logs with chain hash integrity
- Agent feedback loop (confidence → auto-proceed / verify / manual)
- Write-through guarantees across all three surfaces
- Docker Compose deployment (single command)
- Landing page + public repository
Phase 2 — Multi-Agent Foundation
Planned — March–April 2026
IAgentAdapterinterface + OpenClaw adapter- Claude Code adapter (Anthropic API)
- Agent Registry (health checks, cost metrics, capability discovery)
- Smart Router v1 (cheapest agent meeting quality threshold)
- Agent RBAC (Owner / Agent / Viewer roles)
- Network policy enforcement (restricted mode)
Phase 3 — Workflow Engine
Planned — April 2026
- Sequential workflow execution with step context passing
- Parallel step execution with result aggregation
- Branching logic (condition-based routing)
- Gemini adapter (Google AI)
- Mistral adapter (optional, cost-optimized)
- Workflow templates for common patterns (Draft-Review-Polish, etc.)
Phase 4 — Skills Marketplace
Planned — April–May 2026
- Skill format specification + parser
- Skill trigger detection and auto-injection
- Core skill library (TDD, debugging, code-review, brainstorming, etc.)
- Skills CLI:
labos skills install/publish/search - Marketplace backend (versioning, ratings, revenue share)
- Community contribution guidelines
Phase 5 — Observability & Scaling
Planned — May 2026
- Prometheus metrics export for all services
- Grafana dashboards (agent health, cost per task, confidence trends)
- Token optimization workflow (pi-autoresearch integration)
- Cost alerts and budget caps per agent
- Horizontal scaling for the Hub and Kanban services
- Redis-backed token cache for high-volume deployments
Phase 6 — Launch
Planned — May–June 2026
- Production-hardened Docker images
- One-click cloud deployment (DigitalOcean, Hetzner)
- SaaS tier (managed ClawPilot, no self-hosting required)
- GitHub Actions CI/CD templates for common workflows
- Product Hunt launch
- Documentation site (docs.dombot.tech)
Out of Scope (v1)
- Fine-tuning or training LLMs
- Real-time voice/video agent interfaces
- Native mobile apps
- Vector database integration (may come in Phase 5+)
Contributing
ClawPilot is MIT licensed. Contributions welcome:
- Agent adapters — wrap your favourite LLM
- Skills — share your workflow patterns
- Bug reports — open an issue on GitHub
- Feature requests — open a discussion
git clone https://github.com/ldom1/dombot-labos cd dombot-labos # See CONTRIBUTING.md for setup instructions