Open-source infrastructure for governable autonomous AI.
Deterministic control · Agent orchestration · Practical automation
Shards Foundation is the open-source engineering home for the wider Shards ecosystem. Our work explores how AI agents can operate usefully without treating model output as permission to act.
The direction is deliberately modular: agent runtimes, orchestration, policy-controlled execution, audit trails, retrieval, and applications that can use those foundations. Projects are at different levels of maturity; a repository's design or roadmap does not imply that its integrations have been deployed.
| Project | Focus | Status / scope |
|---|---|---|
| Shards-Kernel | Python package providing deterministic execution decisions, scoped permits and hash-chained audit evidence. | Published on PyPI; local reference operations, not a general-purpose OS or sandbox. |
| Shards Agent Core | Modular Python agent runtime, specialist agents and workflow coordination. | Framework and integration work. |
| Shards Swarm Agent | Multi-agent registration, task coordination, workflow and monitoring prototype. | Alpha; not production-ready. |
| Super-Intelligence Agent | Multi-provider agent framework experiments and local API scaffold. | Experimental; not a deployed service. |
| AgentMesh | Architecture for connecting agent swarms and messaging-platform bot networks. | Integration-oriented project. |
| Project | Focus |
|---|---|
| Omni-LLM Orchestrator | Intent-aware multi-model routing, weighted response synthesis and chat UI. |
| Shards RAG | Enterprise document retrieval and AI knowledge-hub application. |
| Workflow Hub | Business operations, connected tools and workflow orchestration. |
| Shards Playbooks | Reusable operational workflows, automation patterns and engineering guidance. |
| Revenue Engine | Early-stage digital-product, payment and fulfilment automation. Alpha, not a live commercial service. |
| Project | Focus |
|---|---|
| Shards Agent Trading Network | Agent profiles, verifiable journals, read-only market data and simulated paper execution. Live trading is disabled. |
| PATENT-OS | AI-supported patent and prior-art research. Not legal advice. |
| Nexus Components Library | React component library and AI-assisted UI playground. |
| Lumina | Headless commerce application and integration patterns. |
| AI Agent Sketchbook | Planning, prototyping and governance-oriented agent workspace. |
Shards-Kernel separates an agent's proposed action from an authorised action. It provides a concrete starting point for exploring fail-closed decisions, permits and audit evidence.
python -m pip install shards-kernel
shards-kernel --version
shards-kernel demo --jsonThe demo runs local reference tools. It does not connect an agent to GitHub, wallets, trading venues or infrastructure by itself. For supported commands and integration boundaries, see the package README and agent quickstart.
Applications & operator interfaces
|
Agent runtimes / orchestration
|
Policy, permissions & audit boundary <-- Shards-Kernel
|
Trusted tool adapters / APIs / data
|
External systems
This is the intended architectural direction, not a claim that every linked repository already integrates with Shards-Kernel. An actual deployment needs its own authentication, isolation, secrets management, monitoring, operational controls and testing.
The repositories are public so developers can inspect the implementation, evaluate design decisions and propose improvements. Start with a project's README, check its licence and contribution guidelines, and open an issue or pull request in that repository.
Shards Foundation · Building useful autonomous systems with explicit boundaries.
