Adopting Avo with AI agents
When you use an AI coding agent — Claude Code, Cursor, Codex, and others — to add analytics, it does two things at once: it decides what to track (the event and property names, descriptions, and structure) and it writes the tracking code. With no shared plan to follow, both drift — names come out inconsistent, events duplicate ones you already have, and the implementation rarely matches what anyone intended.
Avo gives that work a governed home. Your agent connects through the Avo MCP and proposes changes against a shared tracking plan with audit rules; every change lands on a branch — never on main — and merging stays a deliberate human step in the Avo web app. Avo governs both sides of what your agent generates: the design (what’s tracked, named to your conventions, without duplicates) and the implementation (code that matches the plan).
Two Avo MCP Agent Skills teach your agent to do this well — one for starting from scratch, one for working within a plan you already have. Pick the page that matches your situation:
Related
- Avo MCP Agent Skills — the
avo-mcpandavo-mcp-new-tracking-planskills that guide your agent through the Avo MCP. - Agentic data design — Avo’s in-app AI suggestions for designing tracking.
- What is a Tracking Plan? — the governance model Avo enforces.
- Avo MCP overview — setup, tools, OAuth, and the branch-write guarantee in full.