Production-grade platform for AI agent memory, context assembly, and secure LLM proxying.
Docs · Benchmarks · Developer guide · Security
Phases 0–2.5 are complete (auth, multi-provider proxy, tokenizer, response pipeline, embedder, MCP skeleton, local LGTM). Phase 3 (Core Memory Substrate) is in progress: schema v2 / HNSW, VectorStore, and the full write pipeline (POST /v1/memories — PII, dedup, conflict, orchestration, multi-label labels[]). Next: Track D read path (semantic search, hot cache). Extraction/context assembly are Phase 3.5; operator platform is Phase 4; intelligence is Phase 4.5; advanced retrieval is Phase 5. See roadmap current state.
- Prerequisites: Docker, Go, Buf, Make, Node 22 / pnpm 9. See TOOLCHAIN.md.
- Full setup and local workflow: DEVELOPMENT_GUIDE.md.
git clone https://github.com/Rick1330/ibex-harness.git
cd ibex-harness
make compose-dev-up
make db-migrate
make db-seed- Architecture: ARCHITECTURE.md
- Environment variables: ENVIRONMENT_VARIABLES.md
- Contributing: CONTRIBUTING.md — see Good first issues for beginner-friendly tasks
- API reference: ibexharness.com/docs/api-reference/chat-completions (REST + gRPC interfaces)
- OpenSSF Best Practices: OPENSSF_BEST_PRACTICES.md