Categories
Developer Tools Knowledge Management RAG Vector Search AI Agent Infrastructure MCP Server Enterprise CollaborationStars
Forks
Watchers
Developer links
AKB
Achieving 98.4% Recall@5 on LongMemEval-S without a reranker, AKB is the agent-native knowledge infrastructure giving Claude Code, Cursor, Windsurf, and any MCP-aware agent persistent organizational memory — a drop-in replacement for Confluence or Notion where AI agents read and write directly rather than scraping human-formatted pages. Documents are stored as Markdown with YAML frontmatter in Git bare repositories (full version history, branching, diff), while PostgreSQL 16 serves as source of truth for chunk text, metadata, and BM25 vocabulary, with dense embeddings handled by a pluggable vector store (pgvector default, Qdrant optional, or managed Seahorse). Agents interact through 20+ MCP tools — akb_put, akb_search, akb_browse, akb_relations, akb_sql, akb_edit, akb_history — performing hybrid dense plus BM25 search fused via Reciprocal Rank Fusion in one call. The URI graph (depends_on, related_to, implements relations) lets agents traverse knowledge connections without external graph databases. Multi-tenant vault isolation is enforced at the PostgreSQL ACL level with per-user roles and SET LOCAL ROLE for SQL sandboxing. Event fanout streams structured events to Redis Streams for external consumers (synthesis bots, digest agents, audit trails), and a hash-chained append-only audit log integrates with enterprise SIEMs. Ships with agent plugins for Claude Code and Codex (session capture, wiki ingest, lifecycle hooks), Kubernetes kustomize manifests, and Open Knowledge Format export. Running on a dedicated VPS on RepoCloud with guaranteed CPU, RAM, and SSD, full root SSH access, and a browser serial console. BSL 1.1 licensed (free under 100 seats, auto-converts to Apache 2.0 after four years).
Benefits
- Agent-Native MCP Knowledge Access
- Twenty-plus MCP tools let AI agents directly read, write, search, and traverse knowledge without human-formatted page scraping or API wrapper translation layers.
- 98.4% Recall Hybrid Search
- Dense semantic embeddings via BGE-M3 fused with BM25 lexical search through Reciprocal Rank Fusion deliver benchmark-leading retrieval accuracy without requiring a reranker.
- Git-Backed Full Version History
- Every document lives in a Git bare repository providing complete version history, diff, branching, and rollback with no additional versioning infrastructure required.
- Enterprise Multi-Tenant Isolation
- PostgreSQL native ACL enforcement with per-user roles and SET LOCAL ROLE sandboxing ensures vault-level data isolation without application-side SQL inspection.
Features
- URI Knowledge Graph
- Explicit depends_on, related_to, and implements relations in YAML frontmatter form a traversable knowledge graph without requiring a separate graph database.
- Pluggable Vector Store
- Choose pgvector (default, zero extra containers), Qdrant (scalable), or Seahorse Cloud (managed) — switch via config with no schema migration.
- Event Stream Fanout
- PostgreSQL events outbox plus optional Redis Streams enable external consumers like synthesis bots, digest agents, and audit pipelines via XREAD consumer groups.
- Agent Lifecycle Plugins
- Claude Code and Codex plugins (akb-wiki, akb-sessions, akb-claude-code) capture session notes, ingest sources, and hook into agent start/compact/end events.
- Hash-Chained Audit Log
- Append-only JSONL audit at the MCP dispatch point with SHA-256 chaining and optional WORM S3 handoff integrates with Splunk, QRadar, or Elastic SIEMs.