Oxagen Docs

Quick Start

Oxagen is the ontology layer for AI agents — a typed, queryable knowledge graph your agents plug into instead of building one.

What is Oxagen?

Oxagen is the ontology layer for AI agents — a typed, queryable knowledge graph that developers and CIOs plug into their agentic stacks instead of rolling their own. You send natural-language events, notes, or structured entities; Oxagen extracts typed nodes and relationships, indexes them for semantic and keyword search, and exposes them over a tight JSON API and the Model Context Protocol (MCP).

Every workspace is an independent, workspace-scoped, deterministic graph. No cross-tenant contamination. Neo4j-backed. MCP-native.

Quick Start

1. Sign up

Create an account at app.oxagen.ai. Signup automatically provisions your workspace — nothing to configure.

Sign in at app.oxagen.ai and open Settings → API Keys to mint a workspace-scoped bearer token. The secret is displayed once at creation — Oxagen stores only a hash. MCP clients wire it up automatically through the installer.

Authorization: Bearer <token>

2. Add something to the graph

curl -X POST https://api.oxagen.ai/v1/ontology/prompt \
  -H "Authorization: Bearer <token>" \
  -H "Content-Type: application/json" \
  -d '{ "prompt": "Met with Sarah Chen (sarah@acme.com), VP Sales at Acme Corp. They want to close a $80k enterprise deal by end of quarter." }'

Oxagen extracts typed nodes (Person, Organization, Deal) and typed edges (works_at, discussed) and returns a count of what was created.

3. Ask a question

curl -X POST https://api.oxagen.ai/v1/ontology/prompt \
  -H "Authorization: Bearer <token>" \
  -H "Content-Type: application/json" \
  -d '{ "prompt": "Which enterprise deals are close to closing?" }'

4. Search semantically

curl -X POST https://api.oxagen.ai/v1/ontology/nodes/search \
  -H "Authorization: Bearer <token>" \
  -H "Content-Type: application/json" \
  -d '{ "q": "fintech contacts I spoke with last month", "top_k": 10 }'

5. Plug in via MCP

Point any MCP-compatible client (Claude Code, Cursor, VS Code, Windsurf, Codex) at Oxagen's MCP endpoint and your agent can read and write the graph directly. The install selector on oxagen.ai hands you a one-line command per client.

Core Concepts

  • Ontology — Typed nodes, typed edges, workspace-scoped graphs
  • Ingestion — Natural-language prompts, structured inputs, and how extraction works
  • Agent Overview — How the agent reaches your workspace from chat, MCP, and the REST API
  • Agent Tools — The canonical capability catalog
  • Artifacts — Documents, slides, sheets, PDF, brand kits, and where generated artifacts get stored
  • Events, triggers, and audits — Ontology events, workspace agent triggers, the audit chain
  • Linear — Ingest issues, projects, milestones, customers, and inferred features
  • GitHub App (private repositories) — Connect private and org-owned repos for code ingestion
  • API Reference — Every public endpoint
  • API Explorer — Interactive reference generated from the OpenAPI spec
  • Security — Workspace isolation, encryption, and privacy controls

On this page