Adobe Analytics MCP server for Claude Code allowing you to query report suites and run freeform reports by metric/dimension/date range via the Adobe Analytics 2.0 API.
Adobe Analytics MCP Server
An MCP server that gives Claude direct access to Adobe Analytics 2.0 — run reports, explore dimensions and metrics, and analyze data conversationally.
Prerequisites
- Python 3.11+ and uv
- R 4.x and the
adobeanalyticsrpackage - An Adobe Analytics OAuth client (Client ID + Secret) and your Company ID
Install R dependencies
install.packages(c("adobeanalyticsr", "jsonlite"))
Setup
-
Clone the repo
git clone <repo-url> cd aa_mcp -
Add to your Claude Code
settings.json— pick one auth method:OAuth (default — interactive browser login, good for personal use):
{ "mcpServers": { "adobe-analytics": { "command": "uv", "args": ["--directory", "/path/to/aa_mcp", "run", "server.py"], "env": { "AW_CLIENT_ID": "your-client-id", "AW_CLIENT_SECRET": "your-client-secret", "AW_COMPANY_ID": "your-company-id" } } } }Server-to-Server / S2S (non-interactive, good for teams and automation):
{ "mcpServers": { "adobe-analytics": { "command": "uv", "args": ["--directory", "/path/to/aa_mcp", "run", "server.py"], "env": { "AW_AUTH_TYPE": "s2s", "AW_AUTH_FILE": "/path/to/adobe-credentials.json", "AW_COMPANY_ID": "your-company-id" } } } }The S2S credentials JSON is downloaded from the Adobe Developer Console: Console → your project → OAuth Server-to-Server → Download JSON
On Windows use backslash paths, e.g.
C:\\Users\\you\\aa_mcp. -
Restart Claude Code — the server will start automatically.
Tools
| Tool | Description |
|------|-------------|
| list_report_suites | List all available report suites (rsid + name) |
| run_adobe_report | Run a freeform report by metrics, dimensions, and date range |
Example prompts
- "What are my report suites?"
- "Show me pageviews and visits by page for the last 30 days"
- "Compare last week's top 5 pages to the week before"
- "What marketing channels drove the most revenue this month?"
Docker alternative
docker build -t aa-mcp .
docker run --rm -i \
-e AW_CLIENT_ID=your-id \
-e AW_CLIENT_SECRET=your-secret \
-e AW_COMPANY_ID=your-company-id \
aa-mcp