Developer reference
Build on AXO
AXO exposes four integration surfaces so both humans and AI agents can operate the platform. Pick the surface that fits how you build – they share one underlying system.
FOR AI AGENTS
MCP server →
Connect Claude or any MCP client to a hosted endpoint and operate AXO through 130+ tools – analytics, segments, variants, placements, triggers, profiles.
Scoped mcpt_ tokens · hosted, no install
FOR YOUR BACKEND
REST API (Pulse v1) →
Read profiles and segment membership from your own systems at decision time. Bearer keys, scopes, cursor pagination, and a published OpenAPI spec.
pls_live_ keys · OpenAPI 3.1
IN THE BROWSER
Tag JS API →
The window.axo API and data-vt-* attributes: identify visitors, fire conversions, pass cart state, and author personalization zones from your markup.
window.axo · data-vt-* attributes
SERVER-SIDE
Ingestion & webhooks →
Send identify / convert / lead events from a server or headless stack, and receive HMAC-signed webhooks for profile, segment, and conversion events.
Public site_id · HMAC-SHA256 webhooks
Authentication at a glance
Generate API tokens in the AXO dashboard and keep them secret. The site_id identifies a site publicly and ships in the browser tag; it is not a secret credential.
| Surface | Credential | Where to get it |
|---|---|---|
| MCP server | mcpt_… (scoped: read/write/admin) | Admin → Connect AI Tools |
| Pulse REST | pls_live_… (scoped keys) | Admin → Pulse |
| Tag / ingestion | Public site_id (client-side) | Admin → Install |
| Webhooks (inbound to you) | HMAC-SHA256 shared secret (X-VibeTag-Signature) | Admin → Integrations |
Machine-readable surfaces
Built for agents and AI answer engines as much as humans:
- /mcp/tools.json – the full MCP tool manifest (names, scopes, input schemas) as JSON.
- /mcp/llms-full.txt – the entire tool + API surface as flat markdown, for LLM ingestion.
- /llms.txt – a compact map of the whole site for AI agents.
/api/pulse/v1/openapi.json– the Pulse REST OpenAPI 3.1 spec.