Abacum MCP: what Abacum says it does, what access it gets, and how to connect it

Abacum says an official MCP connection lets Claude, ChatGPT and Gemini query Abacum data. URL, auth and write access are undocumented. Verified 21 Oct 2026.

Abacum describes its MCP connection only on marketing pages. It publishes no developer docs, server URL, auth method or tool list, so check the details with Abacum before you plan a rollout.

Official MCP · read/write not documented

Hosting: Not documented. An Abacum product image shows its Connections panel linking an MCP server with AI tools, which points to a hosted connection; no server URL is published.
Auth: Not documented by the vendor.
Read / write: Not documented. Abacum describes asking questions about and querying Abacum data; it doesn't say whether the AI can change anything.
Works with: Claude, ChatGPT and Gemini, per Abacum; a product image also shows Copilot. Abacum adds "other LLMs" without naming them.
Plan requirement: Not documented. Abacum is sold by demo and quote. Claude custom connectors work on every Claude plan, and Free is limited to one.
Last changed: 2026-10: MCP connection described on Abacum's Intelligence pages
Verification: Verified 21 Oct 2026 by theFinanceStack · confidence: medium

What can an agent do with Abacum through MCP?

Per Abacum, a user can ask questions about Abacum's financial planning data from Claude, ChatGPT or Gemini. The same role-based access set up in Abacum applies. Abacum publishes nothing more specific.

Abacum mentions MCP on two pages: the Abacum Intelligence product page and its landing-page version, which carries a short FAQ on how the MCP integration works. Both are marketing pages.

What Abacum doesn't publish

Abacum publishes no tool list, tool count, API reference, rate limits or example prompts, so this section is limited to what those pages say.

Abacum Intelligence and MCP

Abacum Intelligence itself is Abacum's in-product AI for planning.
The MCP connection is for teams who would rather ask from an AI assistant they already run.
Abacum says its in-product AI Modeler builds, edits or explains models and shows what will change before it changes it.
Abacum doesn't document those edit features as reachable through MCP.

Can

Answer questions about Abacum financial data from Claude, ChatGPT or Gemini, per Abacum
Query Abacum data directly from those tools instead of exporting it, per Abacum's FAQ
Work within the role-based access controls and data governance configured in Abacum, per Abacum
Be linked to the AI tools an organisation uses from Abacum's Connections panel, per an Abacum product image

Cannot

Abacum doesn't document any write, update or approval actions through MCP, so don't assume the AI can change plans, forecasts or scenarios
Abacum doesn't publish a server URL or self-serve setup steps, so you can't connect it without Abacum's help today
Abacum lists role-based access controls, version history and approval workflows as its governance controls, and doesn't say whether these apply to queries made through MCP

What access does it get?

Abacum says the AI works within the same permission structure as your team in Abacum. It doesn't publish OAuth scopes, a sign-in flow or which Abacum user the connection runs as.

Where	Scopes and permissions (as documented)	Admin requirement
Abacum MCP connection (Claude, ChatGPT, Gemini)	Not documented by the vendor. Abacum says queries run with the role-based access controls and data governance set up in Abacum, and that no data leaves your permission boundaries.	Abacum doesn't document who can link the connection; its image shows it in Abacum's Connections panel. On Claude Team, an Owner or Primary Owner adds custom connectors; on Enterprise, so can people with a custom role that includes Manage access to Libraries, and members then connect individually. In ChatGPT Business/Enterprise an admin must allow the app or developer mode.

How do I connect it?

Abacum doesn't publish setup steps or a server URL. Start from Abacum's Connections panel or ask your Abacum contact for the endpoint. Then add it to your AI client with that client's standard steps.

In Abacum, open the Connections panel and link the MCP server to your AI tool, as Abacum's product image shows; Abacum doesn't document this step in writing, so confirm it with Abacum.
Get the MCP server URL from Abacum; it isn't published.
Ask Abacum whether each person signs in with their own Abacum user or everyone shares one login or API key. Anthropic says a shared credential gives everyone who uses the connector that credential's access.
Pro or Max: in Claude.ai or Claude Desktop, open Customize › Connectors, choose + Add › Add custom connector and paste the URL. Team or Enterprise: the admin adds it under Organization settings › Connectors › Add › Custom › Web instead.
Choose Connect and complete whatever sign-in Abacum's server asks for; Abacum doesn't document the auth method.
Team: an Owner or Primary Owner adds the custom connector first. Enterprise: an Owner, Primary Owner or someone with a custom role that includes Manage access to Libraries adds it. Members then find it under Customize › Connectors and connect individually.

Abacum names Claude but documents no setup steps; these are Claude's standard custom connector steps. Abacum isn't in Claude's connector directory as of 21 Oct 2026 (claude.com/connectors/abacum returns 404).

What should a finance team watch for?

Abacum publishes very little. Before connecting, get the endpoint, auth method, read/write behaviour and limits from Abacum in writing.

Abacum publishes no server URL, auth flow, scopes, tool list, rate limits, row caps or pricing for the MCP connection; ask for them before a rollout.
Read/write is undocumented. Until Abacum confirms the connection is read-only, treat it as able to change data and keep human approval on in your AI client.
Ask Abacum whether the connection signs each person in with their own Abacum user or runs everyone through one shared login or API key. Anthropic says a shared credential gives every user that credential's access, which would bypass the role-based limits below.
Least privilege comes from the Abacum user and role. Per Abacum the AI sees what that role sees, so connect users whose access fits the questions they will ask.
On training, Abacum's FAQ says customer financial data is "never used to train Abacum's AI models or any third-party LLM"; your AI provider's terms govern what happens in the chat.
Abacum's FAQ says customer data is isolated, encrypted and governed by SOC 2 Type II, ISO 27001 and GDPR standards. Abacum's security page doesn't mention MCP or AI connections, so ask Abacum to confirm those attestations cover the MCP connection.
Model names, account labels and commentary are free text the agent reads; an injected instruction there could push write-capable tools in the same chat to change or send data, so keep approval on for those.
Check AI answers against the Abacum model before they go into a forecast, budget or board pack.

How could finance teams and accounting firms use it?

Abacum documents only question-asking. That suits analysis and reporting on Abacum plans and actuals, with every change to the model still made by a person in Abacum.

Inside a company

Ask which departments are over budget this quarter and get a short summary from the Abacum data your role can see.
Draft budget against actuals commentary for a monthly pack, for the FP&A lead to check against Abacum before it goes out.
Let a RevOps or department lead ask their own headcount or spend questions within the access Abacum already gives them.
Combine Abacum figures with another connected source in the same chat to prepare questions for a forecast review.

At an accounting firm

Answer a client CFO's planning question during a call from their Abacum data, using an Abacum user limited to that client.
Prepare a first draft of variance commentary for a client board pack, then review it against the client's Abacum reports.
Before an FP&A engagement, ask a client's Abacum data for budget against actuals by department to see where the review should start.

Any change to plans, scenarios or the general ledger stays with a person who approves it in Abacum or the source system.

Want Abacum connected to Claude for your team?

We configure scoped connections, with a review step before anything changes your books, for a fixed fee.

Book a 30-minute setup call

Change log

Date	Change	Source
2026-10-21	Re-verified: still marketing-level only; no public server URL, auth, tool list or Claude directory listing	www.abacum.ai/lp/intelligence
2026-10	Abacum's Intelligence pages describe connecting via MCP to Claude, ChatGPT and Gemini, with an FAQ on how the integration works	www.abacum.ai/product/intelligence

FAQ

Is this official?

Abacum itself says it connects via MCP, so in that sense, yes. It isn't listed in Claude's connector directory, and Abacum publishes no developer documentation for it.

Is it free?

Abacum doesn't state a price or plan requirement for the MCP connection. Abacum itself is sold by demo and quote. Claude custom connectors work on every Claude plan, and Free is limited to one.

Does it change data in Abacum?

Abacum doesn't say. It describes asking questions and querying data. It doesn't describe writing, and it hasn't documented the connection as read-only, so confirm with Abacum.

Is my data used to train models?

Abacum says your financial data is not used to train Abacum's models or any third-party LLM. Your AI provider's terms govern what happens inside the chat, so check both.

Can my accountant use it on my behalf?

Abacum doesn't address this. Access follows Abacum's role-based controls, so an outside adviser would likely need their own Abacum user with a suitable role; ask Abacum.

Something wrong or out of date? Tell us and we'll email you when it's fixed.

Sources

Verified by theFinanceStack, 21 Oct 2026. Vendor list on the CFO Software Market Map courtesy of Jason Staats; AI connectivity data by theFinanceStack. Not tax, compliance or financial advice.