Athena-Public
Turn stateless LLMs into a stateful, compounding AI agent. Local-first persistent memory, quantitative decision screens, and governance harness across Claude Co
What is Athena-Public?
Turn stateless LLMs into a stateful, compounding AI agent. Local-first persistent memory, quantitative decision screens, and governance harness across Claude Code, Antigravity, Cursor, and VS Code. Own the state. Rent the intelligence.
Overview of Athena-Public
What it is
Athena is an open-source SDK that turns stateless large language models (LLMs) into stateful, compounding AI agents. It provides local-first persistent memory, quantitative decision screens, and a governance harness. Memory is stored as plain Markdown and JSONL files on the user's disk, making it portable across multiple IDEs including Claude Code, Antigravity, Cursor, Gemini CLI, and VS Code. The project separates intelligence (the rented model) from state (the user's sovereign disk).
Key features
- Persistent memory and hybrid retrieval: memories are stored and surfaced when relevant using hybrid RAG with document-level RRF fusion and cross-encoder reranking.
- Problem framing and Value of Information (VOI) ledger: pre-registers falsifiable hypotheses, commitment gates (R1–R13), and Brier calibration scoring.
- Quantitative decision engine: deterministic evaluators for Expected Economic Value (EEV), One-Third Kelly, Absorbing-Barrier Ruin, MCDA with ruin vetoes, Decision Monte Carlo, Scenario Robustness, and Barbell strategies. All decisions produce SHA-256 receipts.
- Adversarial wargaming engine: two-turn sequential Blue vs. Red game trees with intra-turn path-dependent trough ruin and tamper-evident run receipts.
- Local analytical SQL: zero-dependency DuckDB engine for profiling and querying large CSV, JSON, and Parquet datasets without exhausting LLM context windows.
- Runtime governance and safety: capability permission gates, crisis safety screen, destructive command blocker, secret egress scanner, and doom-loop circuit breaker.
Use cases
Athena is designed for developers and teams using AI coding assistants who need persistent memory, structured problem framing, quantitative decision screening, and governance. It is suitable for daily engineering and analysis tasks as well as high-stakes strategy and architecture work. The three boot modes (Lightweight, Standard, Deep) allow users to choose the appropriate level of context and token cost.
Getting started
Install the minimal version:
pip install athena
For full installation with cloud sync and cross-encoder reranking:
pip install -e ".[full]"
See the repository README for Windows setup, advanced configuration, and Supabase setup.
License and pricing
Athena is released under the MIT License. It is open source and free to use.
Based on the project's public README and repository metadata.