# Avo [Health: Active]

**Category:** 📊 Data Platforms  
**Repository:** https://github.com/avohq/avo-mcp  
**GitHub Stars:** 1  
**Views:** 63  
**Installs:** 0  
**Upvotes:** 1  
**Directory Page:** https://allmcps.com/mcp/avo

## Description
Define, ship & query your analytics tracking from one source of truth, trusted by humans and agents.

## Claude Desktop Quick Installation
Remote MCP endpoint (confidence: high). Install path detected from listing signals. Add as a URL/SSE server in your client:

```json
"mcpServers": {
  "avo": {
    "url": "https://mcp.avo.app/mcp"
  }
}
```

## Documentation

## What Avo MCP server does

The Avo MCP server lets AI agents work with an Avo tracking plan through MCP. Agents can look up events, properties, and metrics by name, ID, meaning, or structural filters, then use the available editing and workflow operations to make changes. The repository also includes guidance for designing instrumentation against an existing plan and for starting a new plan in an empty or nearly empty workspace.

The included skills support two workflows. `data-designer` helps teams explore an existing plan, design tracking for a feature or PRD, review branches, and make small branch edits. `data-designer-new-plan` guides planning discussions, naming conventions, initial categories, and starter events and properties before creating a branch.

## How it works

The Avo MCP server is hosted remotely at `https://mcp.avo.app/mcp` and uses OAuth for authentication. MCP clients connect to that endpoint over HTTP. The repository's skills provide instructions and bundled resources that Claude can load when relevant; they are separate from the hosted server's tool implementation.

Changes are branch-based. Writes do not go directly to `main`, and the MCP server does not merge branches. A human must merge a completed branch into `main` in the Avo web application. The repository is the source for the included skills and plugin configuration, with no build step required there.

## Setup and configuration

Add the remote endpoint to an MCP-compatible client. Claude Code can register it with:

```bash
claude mcp add --transport http avo https://mcp.avo.app/mcp
```

Claude Desktop or Claude.ai can use a custom connector with the same URL. Cursor and other clients that support remote streamable HTTP servers can use an MCP configuration such as:

```json
{
  "mcpServers": {
    "avo": { "type": "http", "url": "https://mcp.avo.app/mcp" }
  }
}
```

Authentication occurs in a browser on first use. The first write also triggers a one-time read-to-write re-authentication prompt. Claude Code users can instead install the repository's `avo` plugin, which registers the endpoint and bundles the skills.

## Tools and capabilities

The Avo MCP server exposes the following documented tool groups:

- `search` for finding tracking-plan items.
- `get` for retrieving items.
- `save_items` for saving edits.
- `workflow` for working through planning or review steps.
- Branch tools for branch-oriented changes.

The skills describe designing new tracking plans, including goals, metrics, funnels, naming conventions, categories, events, user properties, system properties, and constraints. They can also be used alongside a connected data MCP for diagnosing tracking gaps; the README names Amplitude, Mixpanel, PostHog, BigQuery, Snowflake, Databricks, and Redshift as examples.

## Limitations and notes

This repository contains the two Agent Skills and the Claude Code plugin manifest and configuration. Its MIT license applies only to those repository files, not to the hosted Avo MCP server, Avo web application, SDKs, or other Avo services.

The server's canonical, current tool reference is maintained in Avo's tools documentation. Merging remains a human action in the Avo web application, so the Avo MCP server is not a replacement for that approval step.

_Full upstream README: https://allmcps.com/mcp/avo/readme_