Context Infrastructure for AI Agents

Persistent memory. Queryable knowledge. One API.

Oxagen is the first ontology-as-a-service platform: a managed knowledge graph that gives AI agents durable memory and structured reasoning context. Connect once over MCP and your agent gains a mind that persists across sessions, tools, and runs.

We are creating the Context Infrastructure category. Stop stuffing prompts — start querying a mind.

Agents forget. Vector stores match keywords without understanding them. Context windows blow up, decay, and lose the thread. Every workflow restarts from zero.

The missing layer is not a better model. It is persistent, structured knowledge that agents can actually reason over.

• Ontology engine. Entities, relationships, provenance, and semantics, modeled as a queryable graph on Postgres.
• MCP server. Drop-in connection for any agent runtime: Claude, Cursor, Copilot, custom stacks.
• Multi-tenant API. Tenants, workspaces, and projects with RBAC, supporting single-user and team configurations.
• Hybrid retrieval. Qdrant for semantic recall, PuppyGraph for graph traversal, joined under one query plane.
• Schema authoring. Design ontologies in code or in the studio, then version them like data.

Running in production on Next.js 15, FastAPI, PostgreSQL (Citus on Crunchy Bridge), Qdrant, PuppyGraph, Anthropic, and Vercel.

pip install oxagen

from oxagen import Workspace

ws = Workspace.connect(api_key="ox_...")

ws.remember("Acme signed a 3-year contract worth $1.2M starting Q3 2026")
ws.query("What is Acme's current commitment?")

# Or wire it straight into an MCP-compatible agent
npx @oxagen/mcp --workspace acme-prod

Oxagen is hiring its Founding Member of Technical Staff. AI-workflow-first, cross-functional, shipping fast. Mail mac@oxagen.ai with whatever best represents your work.

• Web: oxagen.ai
• Docs: docs.oxagen.ai
• Contact: hello@oxagen.ai
• Founder: Mac Anderson