.png)
3 months ago
2
Where Intelligence Lives.
CORE is a self-governing, constitutionally aligned AI development framework that can plan, write, validate, and evolve software systems — autonomously and safely. It is designed for environments where trust, traceability, and governance matter.
The core self-governance and constitutional amendment loop is complete and stable. The system can successfully audit and modify its own constitution in a secure, principled way using a human-in-the-loop, cryptographically signed approval process.
The next major phase of development, as outlined in our Strategic Plan, is to build the agent capabilities that will allow CORE to generate and manage entirely new applications based on user intent.
We are making the project public now to invite collaboration on this foundational architecture.
- 🛍️ A system that evolves itself through declared intent, not hidden assumptions
- 🛡️ A platform that enforces constitutional rules, domain boundaries, and safety policies
- 🧹 A modular agent architecture with a clear separation of concerns
- 📜 A framework where every decision is documented, reversible, and introspectable
| .intent/ | The "mind" of CORE: contains the constitution, policies, capability maps, and self-knowledge. |
| ConstitutionalAuditor | The system's "immune system," which continuously verifies that the code aligns with the constitution. |
| PlannerAgent | The primary AI agent that decomposes high-level goals into executable plans. |
| core-admin CLI | The secure, human-in-the-loop tool for ratifying constitutional changes. |
| Canary Check | A safety mechanism where proposed changes are audited in an isolated "what-if" environment before being applied. |
| Knowledge Graph | Tracks symbols, roles, capabilities, and relationships across the codebase. |
| Git & Rollback | All changes are version-controlled, and the system is designed for safe rollback of invalid modifications. |
-
Clone the repository:
git clone https://github.com/YOUR_USERNAME/core.git cd core -
Install dependencies:
-
Set up your environment:
cp .env.example .env # Edit .env with your API keys and paths. See .intent/config/runtime_requirements.yaml for all required variables. -
Run the self-audit: Before running the system, verify that your local setup is constitutionally valid.
python -m src.core.capabilitiesThe expected output is ✅ ALL CHECKS PASSED (0 warnings).
-
Launch the CORE server:
- Plans improvements using AI agents.
- Generates code, tests, and docstrings.
- Performs self-audits to ensure constitutional alignment.
- Enforces a secure, human-approved process for self-modification.
- Self-corrects when validation fails.
- Logs every step for transparency.
Unlike most auto-dev tools, CORE:
- Enforces separation of duties between agents and roles.
- Tracks capabilities per function/class with # CAPABILITY: tags.
- Aligns all actions to a declared and auditable constitution.
- Operates with rollback, review, and cryptographic validation by default.
- Supports critical infrastructure and governance-heavy use cases.
We welcome contributions from:
- AI engineers
- DevOps/GitOps pros
- Policy designers
- Governance/compliance experts
👉 See CONTRIBUTING.md to get started. 👉 Check out our Project Roadmap to see where we're headed.
This project is licensed under the MIT License. See the LICENSE file for details.
CORE was born from a simple but powerful idea: "Software should not only work — it should know why it works, and who it’s working for."
