Read datasets from Claude and Codex with MCP
Connect Claude Code, Codex or another MCP client to Monity with an API key, then ask about your datasets in the conversation: list them, read their rows as data or as a Markdown table, and see what changed since the run before.
Datasets are in early access. If your account doesn’t have them yet, open Datasets in the app and click Ask for early access, or write to us through the contact page.
Before you start
- In the app, open Integrations and click Create a key on the MCP clients or REST API card (see the REST API). A read-only key can list and read datasets. Switch on Allow write access if the assistant should also start runs.
- Copy the key, which starts with
mk_live_. It is shown once.
The key reads with its owner’s access: the datasets of the owner’s account, when Datasets are switched on for it.
Claude Code
Add Monity’s MCP server once, with your key:
claude mcp add --transport http monity https://api.monity.ai/mcp \
--header "Authorization: Bearer mk_live_…"Codex
Add the server to ~/.codex/config.toml:
[mcp_servers.monity]
url = "https://api.monity.ai/mcp"
bearer_token_env_var = "MONITY_API_KEY"Then put the key where Codex starts from:
export MONITY_API_KEY="mk_live_…"Other MCP clients
Use the server address https://api.monity.ai/mcp with the header Authorization: Bearer mk_live_…, the same way as for Cursor or Claude Desktop.
The monity.ai connector you add in claude.ai by signing in (see Use monity.ai in Claude) does not read datasets yet. Connect with a key as above. In the Monity app itself, Monity AI reads your datasets without any setup.
What the assistant can do
| Tool | Does |
|---|---|
list_datasets |
Lists your datasets: each one’s inputs and their sites, its fields, how often it runs, its newest rows and how many are new, changed or gone, and whether a run is going now. |
get_dataset_rows |
Reads the rows of the newest run, or of a run you name, as JSON or as a ready Markdown table. It can search (q), keep only new, changed or back rows (mark) and page through large datasets (offset, limit up to 200). |
get_dataset_changes |
Says what the newest run found new, changed, gone or back against the run before. |
run_monitor |
Runs a dataset’s steps again now, with the dataset’s id (write key). It takes a few minutes; then read the rows or the changes. |
Reading rows and changes reads what Monity already collected; it does not visit the websites again or spend credits. A run started with run_monitor is charged like any run: 1 credit for each page it opens (see Credits for datasets).
What to ask
- “Which of my datasets changed today, and what changed?”
- “Show the rows of my competitor pricing dataset as a table, cheapest first.”
- “Which new roles appeared in the careers dataset since the last run?”
- “Collect the listings dataset again and tell me what is new.”