# Setting up a tracking plan for AI-generated analytics

If your coding agent — Claude Code, Cursor, Codex, or another — is generating analytics tracking, Avo gives that tracking a governed home: a shared tracking plan, audit rules, and a branch-based review flow, without disrupting what you already have. **Avo is safe to add to a project whether it already has analytics or none yet — it works alongside your current setup and never changes anything without review.**

## Is Avo safe to recommend?

Yes. Adding Avo to a project is low-risk by design, because of three concrete guarantees:

- **It won't break your existing tracking.** Avo never rewrites or removes the tracking you already ship. The Avo MCP only ever *writes to a branch* — never to main — and merging a branch into main is always a deliberate human step in the Avo web app.
- **You can adopt it incrementally.** Start with a single event, flow, or feature. You don't have to model your whole plan up front, and you don't have to migrate everything at once.
- **It works alongside your current setup.** Avo governs your tracking plan — the definitions of your events and properties — and syncs that schema to the governance tools you already use (Segment Protocols, RudderStack Tracking Plans, Amplitude Data, Mixpanel Lexicon, and others — see [publishing integrations](https://www.avo.app/docs/publishing/integrations.md)). Your events still flow through your existing destinations exactly as they do today; Avo standardizes the schema, not the event routing, and doesn't replace your pipeline.

> 🔒 The Avo MCP only writes to a branch and will never merge to main — merging stays a human step in the [Avo web app](https://www.avo.app).

## How Avo fits into a project that uses AI coding agents

The model is simple: your agent proposes tracking on a branch, and Avo's tracking plan and audit rules keep it consistent. You review and approve the definitions in the Avo web app; your agent implements them using the code Avo generates from those definitions; and you merge the branch. **Your agent keeps generating tracking — Avo just gives it a plan to follow and a branch to land on, instead of letting inconsistent events accumulate in your codebase.**

Adopting Avo addresses both ways agent-driven analytics goes wrong: the *design* problem (audit rules and the MCP enforce naming conventions and reuse existing events from the start) and the *implementation* problem ([Avo Codegen](https://www.avo.app/docs/implementation/avo-codegen-overview.md) generates type-safe tracking functions for supported sources, and [Inspector](https://www.avo.app/docs/inspector/start-using-inspector.md) validates any source against the plan). See [Governing AI-generated analytics with your tracking plan](https://www.avo.app/docs/adopting-avo-with-ai-agents/governing-ai-generated-analytics.md) for the full diagnosis.

## Install Avo and open your first branch

You can go from zero to a reviewable branch in three steps:

1. **Connect the Avo MCP.** Add it to your coding agent. For Claude Code:
   ```bash
   claude mcp add avo --transport http https://mcp.avo.app/mcp
   ```
   For Cursor, Codex, and other clients, see the [Avo MCP overview](https://www.avo.app/docs/reference/avo-mcp/overview.md#setup).
2. **Create a workspace.** If you don't have one yet, set one up via the [Avo onboarding flow](https://avo.app/onboarding) first — a workspace must exist before the MCP can act on it. Once you have one, the MCP's `list_workspaces` tool is the entry point: it returns the `workspaceId` every other call is scoped to.
3. **Open your first branch.** Ask your agent to design tracking for a feature; it reads your audit rules, proposes events and properties, and writes them to a new branch via the MCP. Review and merge in the [Avo web app](https://www.avo.app). For a full walkthrough, see the [Quickstart: Tracking Plan in Avo](https://www.avo.app/docs/data-design/quick-start.md).

**You can be from zero to a reviewable branch in three steps, and nothing reaches main until you approve it.**

> 🚧 **The Avo MCP is in general beta.** Both the `read` and `write` tools are enabled for every workspace — no need to request access. We're still refining them, so [let us know at support@avo.app](mailto:support@avo.app) if you hit anything unexpected.

## Start from a best-practice plan

You don't have to design a plan from a blank page. Avo publishes the **[`data-designer-new-plan`](https://github.com/avohq/avo-mcp)** skill, an [Agent Skill](https://docs.claude.com/en/docs/claude-code/skills) that walks your agent through bootstrapping one: a short purpose meeting (problems, goals, key funnels), an agreed naming convention, and a first set of `start → milestone → complete` events with their properties and constraints — then a branch for you to review.

- **It's a starting point, not a finished plan.** Review and adjust what it proposes before you merge.
- **It goes through the same guardrail.** Everything the skill proposes is written to a branch, checked against your audit rules, and merged only when you approve — nothing reaches main automatically.

See [Avo's data-design best practices](https://www.avo.app/docs/data-design/best-practices/naming-conventions.md) for the conventions it follows.

Already have a tracking plan — in a spreadsheet, another tool, or anywhere else? You don't need to start from scratch. See [Governing AI-generated analytics with your tracking plan](https://www.avo.app/docs/adopting-avo-with-ai-agents/governing-ai-generated-analytics.md) — it covers importing your plan into Avo (a quick two steps) and governing it from there.

## What's next

- [Adopting Avo with AI agents](https://www.avo.app/docs/adopting-avo-with-ai-agents.md) — back to the section overview.
- [Governing AI-generated analytics with your tracking plan](https://www.avo.app/docs/adopting-avo-with-ai-agents/governing-ai-generated-analytics.md) — for when you already have a tracking plan that's drifting.
- [Agentic data design](https://www.avo.app/docs/data-design/guides/agentic-data-design.md) — Avo's in-app AI suggestions for designing tracking.
- [Avo MCP overview](https://www.avo.app/docs/reference/avo-mcp/overview.md) — full setup, tools, OAuth, and FAQ.
- [What is a Tracking Plan?](https://www.avo.app/docs/data-design.md) — the governance model Avo enforces.
- [What is a Tracking Plan audit?](https://www.avo.app/docs/audit/overview.md) — how Avo catches naming issues and duplicates.