Privacy-first, local-first memory engine for AI research.
thecontextcache is a local-first, zero-knowledge memory engine that ingests documents, extracts knowledge quads (subject, predicate, object, context), and returns explainable answers with citations and cryptographic audit trails.
It is designed for researchers, students, and power users who need traceable, privacy-preserving AI answers without relying on centralized SaaS platforms.
- Backend → Python 3.13 · FastAPI · PostgreSQL (pgvector)
- Frontend → Next.js (App Router) · TypeScript · Tailwind CSS · Cytoscape.js
- Security → XChaCha20-Poly1305 · Ed25519 signatures · Argon2id KDF · BLAKE3 hash chains
- Infra → Cloudflare Pages (frontend) · Google Cloud Run (backend) · Neon Postgres · Upstash Redis
- Docs → Mintlify
- ✅ Phase 1 Complete → Core models, crypto utilities, storage adapters
- 🚧 Phase 2 In Progress → MCP servers (docs, extractor, memory, audit, policy-gate)
- 📅 Upcoming → Frontend integration (Dashboard, Graph, Explain panel), test suites (pytest, Schemathesis, Playwright, k6), Mintlify docs
- contextcache → Main project repository
- Website → thecontextcache.com
- Social → Bluesky
- Email → thecontextcache@gmail.com
We welcome contributions! Please see CONTRIBUTING.md and join the conversation in GitHub Discussions.
Dual-licensed under:
- Apache 2.0 → SDKs, client libraries, CLI tools, documentation
- PolyForm Noncommercial 1.0.0 → server code (backend, analyzers, MCP servers, adapters)
Trademark Notice: thecontextcache™ name and logo are reserved as trademarks of the maintainers.