Monte Carlo
by Monte Carlo
Triage data-quality alerts, trace lineage, read query history and create monitors from your AI assistant. 25 tools, OAuth sign-in, and a documented read-only role.
Verified connector
Listed by Anthropic as a partner connector in its Connectors Directory.
Connection checked by Agentman on .
Anthropic states this reflects the level of review a connector received, not a security audit.
Connect Monte Carlo via MCP
https://mcp.getmontecarlo.com/mcpWorks in any MCP-compatible client. In Agentman Studio it is one click — no config file to edit.
Monte Carlo Tools & Capabilities (25)
Tool names from Anthropic's directory listing. This server requires sign-in, so we could not read tool descriptions or parameter schemas.
Limits
- It cannot delete anything. No
delete,remove,archive,muteorsilenceverb appears in the 25 listed names, the 14 named in Monte Carlo's docs, or the 67 in its Agent Toolkit. - It cannot retune detection. Monte Carlo states alert statuses give no feedback to the threshold models; tuning runs through a separate documented process.
- It cannot reach the warehouse directly. Monte Carlo observes your data stack. The connector reads Monte Carlo's metadata, alerts and query history, not the tables themselves.
- Role.
mcp/accessis required to connect at all. Writes additionally needmcp/edit, which the Viewer role lacks. Monte Carlo's MCP page states Editor or above is required, contradicting its own IAM table. - Toolsets.
extendedandagent_observabilitytools are unavailable unless the client sends thex-mcd-toolsetheader. - SaaS only. Every documented path targets Monte Carlo's hosted service at
mcp.getmontecarlo.com. The root resource descriptor names a regional host,mcp.us1.getmontecarlo.com, indicating regional deployments of the same SaaS. Monte Carlo documents no self-hosted MCP Server.
Frequently asked questions
Connection won't start / shows an error
Monte Carlo's reference page names two causes. Verify the endpoint URL, because the key-based mcp-remote setup expects a trailing slash rather than the bare path used by OAuth clients. Then confirm npx is on your PATH, since the bridge auto-installs only when the -y flag can reach npm.
401 / Unauthorized
Check the credential itself before the client. Monte Carlo tells you to re-check the Key id and Key secret, confirm the headers are passed in one of the two supported forms, and verify the key type is MCP server. Standard Monte Carlo API keys will not work with the MCP Server at all.
HTTP 403: Invalid OAuth error response
Usually an outdated MCP client rather than a permissions problem. Monte Carlo's advice is to upgrade to the latest version. If your console shows an empty Authorization header, the token is not being passed, which the vendor attributes to older client versions that silently omit custom headers.
Which Monte Carlo role can use the MCP server without writing anything?
Viewer. Monte Carlo's built-in roles reference grants Viewer the mcp/access permission for read-only MCP tools, and it is the only one of the six built-in roles with no MCP Edit row. Responder adds write access while staying read-only elsewhere, which suits alert triage.
Can the Monte Carlo MCP server resolve or close a data-quality alert?
It can set the status, which is how Monte Carlo records resolution. Statuses run from Acknowledged through Work in progress, Fixed, Expected and No action needed. Monte Carlo states plainly that alert statuses do not feed back into the models generating thresholds, so a status change never retunes detection.
Does the Monte Carlo MCP server expose SQL written by our engineers?
Yes. getQueriesForTable returns query history, which Monte Carlo's toolkit documents as read and write queries against a table. The Performance documentation shows query text attributed to users and tools such as Looker, retained for the most recent four weeks, so business logic and literal values can appear.
Can you restrict which networks reach the Monte Carlo MCP server?
Yes, by IP address. Monte Carlo publishes a dedicated MCP scope in its Network Access Control Lists covering the mcp.getmontecarlo.com/mcp endpoint. Only requests from allowlisted addresses are permitted once configured, managed in Settings under Network access controls or through the GraphQL API.
Use in Agentman
Connect once and your agents call these tools on their own — on a schedule, in a workflow, with nobody at the keyboard.
Open in Agentman StudioServer Info
- Category
- Data & Research
- Developer
- Monte Carlo
- Tools
- 25
- Domain
- mcp.getmontecarlo.com
Using Claude Desktop or another MCP client? Setup docs — the connection URL above works anywhere.