Changelog

PostgreSQL tsvector full-text search combined with Elasticsearch BM25 scoring, unified via Reciprocal Rank Fusion (RRF). Queries hit both engines and results are merged for higher relevance.

Semantic search powered by pgvector with HNSW indexing (768-dimensional vectors). Supports multiple embedding providers: Gemini, OpenAI, Voyage, and Jina.

Full tool suite for AI agents via the Model Context Protocol:

Works with any MCP-compatible AI tool. Tested and documented for Claude Code, Cursor, Codex, Gemini CLI, Windsurf, and Cline.

Automatic memory consolidation via vector clustering and LLM synthesis. Groups related context entries using embedding similarity, then generates consolidated summaries with OpenAI gpt-4.1-mini. Reduces noise, surfaces patterns.

Full multi-tenant architecture with JWT and API key authentication, team management with role-based access (owner, admin, member), and Row-Level Security isolation at the database layer.

React-based dashboard for managing your context. Includes a context editor, Dream UI for memory consolidation, activity logs, and team management.

AES-256 content encryption with tenant-managed keys, available on the Pro plan. Content is encrypted before storage — Henia cannot read your context content.

Single HTTP endpoint for MCP tool execution. No stdio servers, no local processes. Connect any MCP client over HTTPS with a Bearer token.