SuperLocalMemory
Information-Geometric Memory for AI Agents
SuperLocalMemory V3 achieves 74.8% on LoCoMo with data staying entirely local — the highest local-first score reported. It replaces cloud LLM dependency with three mathematical techniques: Fisher-Rao geodesic retrieval, sheaf cohomology for consistency, and Riemannian Langevin dynamics for lifecycle. EU AI Act compliant by architecture in Mode A.
The Problem
Every AI agent memory system in production relies on cosine similarity over embeddings. At scale, this fails: flat similarity loses discriminative power, pairwise contradiction checking grows O(n²), and hardcoded lifecycle thresholds break on non-average workloads. Cloud-dependent architectures compound this with data residency risk and an August 2026 EU AI Act compliance deadline that engineering alone cannot resolve.
Key Capabilities
Fisher-Rao Geodesic Retrieval
Models each memory as a Gaussian distribution rather than a point. Distance is computed on the statistical manifold using the natural Fisher-Rao metric — confidence-weighted, improving with use.
Sheaf Cohomology Consistency
Detects contradictions in the knowledge graph algebraically via H¹(G, F). Catches global inconsistencies that pairwise checking cannot find, without O(n²) cost.
Riemannian Langevin Lifecycle
Replaces hardcoded thresholds with stochastic gradient flow on the Poincaré ball. Memory states self-organize to a provably optimal distribution based on actual usage patterns.
Three Operating Modes
Mode A: zero cloud, EU AI Act compliant, 74.8% LoCoMo. Mode B: Mode A + local Ollama LLM. Mode C: cloud LLM synthesis, 87.7% LoCoMo. Switch anytime.
17+ Tool Integrations via MCP
Works with Claude Code, Cursor, VS Code Copilot, Windsurf, ChatGPT Desktop, Gemini CLI, and over a dozen other AI tools out of the box via the Model Context Protocol.
SkillFortify
Supply Chain Security for AI Agent Skills
Static analysis, behavioral sandboxing, and cryptographic attestation for the AI agent skill ecosystem. Detects malicious skills before they execute.
AgentAssert
Design-by-Contract for AI Agents
Formal specification and runtime enforcement of behavioral contracts for autonomous AI agents. Prevents drift, ensures compliance, enables composition.
AgentAssay
Regression Testing for Non-Deterministic AI Agents
Token-efficient stochastic behavioral testing framework purpose-built for non-deterministic AI agent workflows.