Skip to main content
Open Source

Policy-Enforced Context Intelligence
for AI Agents

MCP-native context layer built on PostgreSQL + pgvector + Apache AGE. Semantic search with access policies. Every query governed by your rules.

Why OneDB

AI agents need context. But context without governance is a liability. OneDB solves both.

Policy-Enforced Retrieval

Every query passes through your access policies. Control what data each agent can see, enforce compliance rules, and maintain data isolation — automatically.

onedb
onedb.query("revenue forecasts") → policy: sales-team-only → scope: Q4-2025, Q1-2026 → result: 12 documents (3 redacted)

MCP-Native Access

Built for the Model Context Protocol from day one. Any MCP-compatible AI tool — Claude Code, Cursor, Windsurf — gets context without custom integrations.

onedb
mcp://onedb/context → semantic search (pgvector) → graph traversal (Apache AGE) → unified retrieval API

Unified Context Layer

One interface for vector search, graph relationships, and structured data. Built on PostgreSQL + pgvector + Apache AGE. No vendor lock-in.

onedb
PostgreSQL → structured data pgvector → semantic embeddings Apache AGE → knowledge graphs OneDB → unified API

Architecture

Ingest. Govern. Retrieve.

Ingestion

Documents, code, knowledge. Chunked, embedded, indexed.

Policy Engine

Access control, scoping rules, compliance enforcement.

MCP Server

Context retrieval for any MCP-compatible AI tool.

Give your agents the context they need

OneDB is open source, runs locally, and integrates with any MCP-compatible AI tool. No cloud dependency.