Step 4: Read from Any AI
Goal: Read your warm query from more than one AI - all from the same source of truth.
Step 4 of 5. Your warm query is reachable by any AI that speaks MCP. This is "bring any AI" - fire us, keep running.
Install the workshop into your agents
The fastest path is fai, which registers the OpenX MCP server with your coding agents in one command - the exact command for your workspace is in the API Explorer, on the MCP tab, ready to copy:
fai mcp install https://openx.fathym.com/oi-api/mcp --name openx --auth "YOUR_TOKEN"
Generate the token under APIs -> API Keys. Browse the tools with fai mcp preview openx. Setting up
a specific client (Claude, Cursor, VS Code, a custom or no-code client)? See the per-client guides under
Bring Any AI.
Try more than one
Now read the same source of truth from different brains - Claude Desktop, claude.ai or ChatGPT over MCP, Cursor, or a local model. Ask each:
"What are the average signups this hour?"
Each one calls the same run_warm_query against the same workshop. The tools available:
| Tool | Does |
|---|---|
list_data_connections | Discover your data sources |
run_warm_query | Run a named query (your sources of truth) |
run_adhoc_warm_query | Try a one-off query |
tail_telemetry | Watch live data |
(Prefer plain HTTP? Every warm query is also a REST endpoint - see the APIs guide.)
The point
One workshop, many brains, one source of truth. No AI is locked in, and the answer is the same everywhere because it comes from the query you crystallized. Next, let those same agents build.