Connect an agent
One MCP server for your whole account. Agents sign in with your datatomcp login; no keys to paste.
MCP server URL
Streamable HTTP with OAuth. Works with any client that supports remote MCP servers.
https://app.datatomcp.com/mcpSetup by client
- Open datatomcp in the Claude directory, or in Claude open Customize → Connectors and find datatomcp in the directory.
- Click Connect, sign in with your datatomcp account and allow access.
- Ask about your data in any chat. Claude Desktop and the mobile apps use the same connector.
On a Team or Enterprise plan, an owner adds it first under Organization settings → Connectors. You can also use + Add → Add custom connector with the URL above.
- In ChatGPT, open Settings → Plugins (Apps on some accounts) and search “datatomcp”.
- Click Install plugin or Connect, then sign in with your datatomcp account and allow access.
- Ask about your data in a chat. ChatGPT calls datatomcp when your question needs it, or mention it by name.
Not listed for you yet? In Plugins, click + → Add custom MCP server and use the URL above with OAuth. Some accounts need developer mode first (Settings → Apps → Advanced settings). On a workspace, an admin may need to allow it.
- Run the command below in your terminal.
- Inside Claude Code, run
/mcp, pick datatomcp and authenticate.
claude mcp add --transport http datatomcp https://app.datatomcp.com/mcp
- Click Add to Cursor below, or open datatomcp on cursor.directory.
- Confirm the install in Cursor, then log in when it asks: that's your datatomcp sign-in.
Prefer the file? Add {"mcpServers":{"datatomcp":{"url":"https://app.datatomcp.com/mcp"}}} to ~/.cursor/mcp.json.
Any client that supports remote MCP servers works. It finds the sign-in from the URL; there is nothing else to configure.
In the agent panel, open … → MCP Servers → Manage MCP Servers → View raw config and add the server. Then in Agent settings → Customizations, click Authenticate, sign in, and paste the code back.
{
"mcpServers": {
"datatomcp": {
"serverUrl": "https://app.datatomcp.com/mcp"
}
}
}In Cascade, open … → Open MCP config file (~/.config/devin/mcp_config.json, or %APPDATA%\devin\mcp_config.json on Windows), add the server, then sign in when the browser opens.
{
"mcpServers": {
"datatomcp": {
"serverUrl": "https://app.datatomcp.com/mcp"
}
}
}Add to opencode.json (project) or ~/.config/opencode/opencode.json. OpenCode asks you to sign in on first use, or run opencode mcp auth datatomcp.
{
"$schema": "https://opencode.ai/config.json",
"mcp": {
"datatomcp": {
"type": "remote",
"url": "https://app.datatomcp.com/mcp"
}
}
}Run both commands; the second opens your browser to sign in.
codex mcp add datatomcp --url https://app.datatomcp.com/mcp codex mcp login datatomcp
Run MCP: Add Server from the Command Palette, choose HTTP, paste the URL and name it datatomcp, then sign in when asked. Or add it to your MCP config by hand:
{
"mcpServers": {
"datatomcp": {
"type": "http",
"url": "https://app.datatomcp.com/mcp"
}
}
}Run the command (-s user makes it available in every project), then /mcp auth datatomcp inside Gemini CLI.
gemini mcp add -s user --transport http datatomcp https://app.datatomcp.com/mcp
Clients that only run local (stdio) servers can use mcp-remote as a bridge. It handles the sign-in.
{
"mcpServers": {
"datatomcp": {
"command": "npx",
"args": [
"-y",
"mcp-remote",
"https://app.datatomcp.com/mcp"
]
}
}
}First question answered Done
Your agent is connected. More to try on sample_sales:
- Using datatomcp, describe the dataset sample_sales and tell me three things worth knowing about it.
- What is the total units by region in sample_sales?
- Show me the 10 most recent rows of sample_sales and tell me which dataset version you used.
aggregate on sample_sales (3 rows) in 231 ms, 1 h ago.Tools your agent gets
Read-only. Every answer carries the dataset version it came from.
list_datasets | Your datasets with row counts and freshness. |
workspace_overview | What it can answer in one call: datasets, what each column means, your definitions, links and example questions. |
describe_dataset | Columns with what each one means, its unit and whether it adds up; your approved definitions and glossary, links to other datasets, data quality, and today's date. |
query_rows | Filtered, sorted rows, up to 1,000 per call. |
aggregate | Totals, counts, averages, min and max, optionally grouped, across linked datasets too. Uses your approved definitions by name; each answer says its unit, how it was counted and any data warning. |
distinct_values | Exact values of a column, so filters match ("France" vs "FR"). |
chart | A chart from a query, one of 14 data charts, with the takeaway as the title. |
Agent access
Each connected agent and API key sees every workspace you're in, unless you untick some here. Agents can also build: propose definitions, links, glossary terms and the business brief (nothing changes until a person approves) and add data from a link. Set an agent to analyze and it only reads.
No agent connected yet. Once you add datatomcp to Claude, ChatGPT or another client, or create an API key, it shows up here.