Datarails FinanceOS

Name: Datarails FinanceOS MCP Server
Availability: InStock
Author: Datarails FinanceOS

HIPAA CompliantSOC2 ReadyISO 27001 Ready

Finance24 tools

Query consolidated actuals, budgets, forecasts and KPIs from your AI assistant. 24 read-only tools, per-user OAuth that inherits your Datarails permissions, and a paid connector subscription.

Verified connector

Listed by Anthropic as a partner connector in its Connectors Directory.

Connection checked by Agentman on 9 September 2026.

Anthropic states this reflects the level of review a connector received, not a security audit.

Connect Datarails FinanceOS via MCP

https://mcp.datarails.com/mcp

Works in any MCP-compatible client. In Agentman Studio it is one click — no config file to edit.

Use in Agentman

Connect Datarails FinanceOS once and your agents call these tools on their own — on a schedule, in a workflow, with nobody at the keyboard.

Open in Agentman Studio

Datarails FinanceOS Tools & Capabilities (24)

get_aggregation_result_by_alias

get_aggregation_result_by_id

get_company_profile_sections

get_connector_status

get_currency_rates

get_data_by_alias

get_data_by_id

get_distinct_values_result_by_alias

get_distinct_values_result_by_id

get_fields_by_id

get_org_users

get_workflow_guide

list_aliased_fields

list_business_metrics

list_company_profile_sections

list_data_models

list_filebox

list_xl_functions

profile_categorical_fields

profile_numeric_fields

start_aggregation_by_alias

start_aggregation_by_id

start_distinct_values_by_alias

start_distinct_values_by_id

Tool names from Anthropic's directory listing. This server requires sign-in, so we could not read tool descriptions or parameter schemas.

Limits

Read-only is the vendor's implementation, not an enforced scope, and Datarails says so. Its documentation states there is no separate read-only mode constraining the connector below the user's Datarails permissions, and that write-capable tools are on the roadmap. The correct control is the person's Datarails permissions.
One scope, and its name says nothing about access. The authorization server advertises scopes_supported: ["datarails"]. No scope names read and none names write, so no consent screen can offer a narrower grant, and no issued token can be audited afterwards for the absence of write access.
The tool inventory can change with a connector version. Datarails states it may add tools over time, including write-capable ones, and that the authoritative list for your organisation at any moment is the one your AI client displays. Review it before and after a connector upgrade. The 24 names here reflect Anthropic's directory listing on 2026-09-09 and Datarails' documentation as updated 2026-08-26.
Per-query audit is incomplete. Datarails documents session-level audit in AI Center as live and tool-call-level audit events from the MCP server as still in development, and advises asking a Customer Success Manager for current status before depending on per-query attribution. Audit record retention for AI Center is described as being finalised.
It is a paid feature with an administrator gate. An active FinanceOS AI connector subscription is required, and a Super Admin must enable the connector per organization. A connector toggled off is blocked at runtime.
Query limits apply. Row-level retrieval is paginated at 100 rows per page by default with offset paging; aggregation results return one row per requested group subject to a response-size limit, and Datarails states a capped response says so and reports the total row count. Service rate limiting also applies. The endpoint's own response headers advertised x-ratelimit-limit: 120 on 2026-09-09; Datarails does not publish the window that number applies to, so treat it as observed rather than as a documented quota.
Processing is US-based today. The MCP service is hosted in the United States. Region-aware routing has no committed delivery date, and Datarails advises raising in-region processing requirements with a Customer Success Manager before rollout.
Not every platform is available. Datarails states Claude and ChatGPT are enabled by default, Lovable and Copilot can be enabled by an administrator, and Gemini is listed as coming soon and cannot yet be enabled.
We read no tool schemas or safety annotations. The endpoint is OAuth-gated, so parameter shapes and per-tool readOnlyHint / destructiveHint annotations were not observable. Every access claim on this page is documentary — Anthropic's directory listing plus Datarails' own documentation — not our observation of the running server.
This page describes what the tools return. It is not financial or accounting advice. No figure, variance or forecast produced through the connector is interpreted here.

Frequently asked questions

Not with the current tool inventory. Datarails documents all 24 published tools as read-only and states none creates, updates or deletes anything, including fileboxes, formulas, models, permissions or users. Datarails is explicit that this is a property of the tools published today rather than an enforced permission scope, and that write-capable tools are on its roadmap.

Yes, and it cannot exceed them. Datarails states every tool call is authorized against the calling user's role and permission groups exactly as in the Datarails web application. Two colleagues connecting the same organization can therefore see different tables. Datarails also notes there is no narrower MCP-only scope that restricts access below a user's existing role.

Yes. Datarails documents the connector as a paid feature requiring an active FinanceOS AI connector subscription for your organization. A Super Admin must also enable it under AI Center then Manage AI Connectors, which is a per-organization action rather than a per-user one. Datarails directs new setups to a Customer Success Manager or its support address.

The connector is most likely not enabled for your organization. Datarails documents get_connector_status as always available for exactly this check, and it reports whether the connector is switched on. If it is off, a Super Admin enables the platform under AI Center then Manage AI Connectors. Datarails also advises confirming the organization holds an active connector subscription.

Only the result of each individual tool call, already filtered by the calling user's permissions. Datarails states there is no bulk export, dataset copy, sync job or background transfer, and that the MCP layer persists no financial records beyond the request. Row-level reads are paginated at 100 rows per page by default, and aggregation responses are capped by a response-size limit.

In the United States, as of Datarails' documentation updated 2026-08-26. Datarails states the MCP service is stateless and stores no customer data, and that your financial data continues to live at rest only in your existing Datarails environment and its region. Region-aware routing behind the same global URL is under evaluation with no committed delivery date.

Only at session level today. Datarails documents a Super Admin audit log in AI Center covering session ID, member, AI platform, timing, duration and status, plus a per-session event timeline. Datarails states that tool-call-level audit events emitted by the MCP server itself, giving per-query attribution, are still in development and should not be assumed available.