Xero's new integration with Anthropic's Claude gives users read-only access to a small set of financial summary tools. It does not provide invoice-level detail or transaction data.
AI adviser Inbal Rodnay highlighted that limitation after testing the setup during a live session with accountants and outlining the tools now available through the connection.
According to Rodnay, an active Xero subscriber can bring selected financial data into Claude conversations. Available data includes revenue and profit, contacts and receivables, financial position, cash position and top customers by revenue.
Her test began with a simple question about the latest invoice in Xero. Claude replied that the available Xero tools were limited to financial summaries and organisation information, and could not pull individual invoices directly.
That prompted a closer look at the connector. Rodnay said the functions exposed through the integration were limited to get cash position, get contacts and receivables, get financial position, get profit and loss, and get top customers by revenue.
There is also no write access, according to Rodnay, meaning users cannot create or amend records in Xero from within Claude. The integration connects to only one Xero organisation at a time, requiring users to disconnect and reconnect to switch between entities.
That design narrows the product's practical audience. Accountants managing several client files may find the one-organisation limit restrictive, while business owners seeking detailed transaction queries may find the current version too narrow for routine bookkeeping tasks.
Tool limits
The rollout points to a more cautious approach to AI access in accounting systems, where software providers must balance convenience against data sensitivity and workflow risk. By exposing only high-level summaries and keeping the service read-only, Xero appears to have set tight boundaries around what Claude can see and do.
Rodnay framed that as a product choice rather than a technical shortcoming.
"The Xero MCP integration is real and it works. But the MCP server Xero built exposes a limited set of tools to the AI. That's not a flaw in the technology. It's a decision Xero made about what to surface," said Inbal Rodnay, AI Adviser.
"And for now, they've chosen high-level summaries, not granular data. And they chose read-only, not every day activity," she added.
Those limits matter because one of the most immediate uses accountants have explored for generative AI is automating repetitive review work. Rodnay described an example from the same session involving a New Zealand accountant who had built a workflow to create a GST review workpaper by taking source data, organising it and flagging issues and recommendations.
In that case, the remaining manual step was downloading reports from Xero. As tested by Rodnay, the current Claude connection does not remove that export step if the user needs underlying records rather than summary figures.
"And it only connects to one Xero org at a time. If you want to switch to another file, you have to disconnect and reconnect," she said.
Microsoft search
Rodnay also looked at options in Microsoft's ecosystem, which is widely used by accounting practices through Copilot and Teams. She said she found a service called Finance Admin for Xero, offering invoice and bill categorisation, approvals in Teams and financial summaries.
That service, however, was a third-party plug-in rather than a tool from Xero or Microsoft. Rodnay said that raised separate questions about supplier identity, data handling and storage before it could be adopted into a firm's workflow.
The comparison underlines a broader issue in the market for AI assistants in finance: even where useful functions exist, they may sit across several vendors, increasing the diligence required from firms handling sensitive client information.
Existing routes
Rodnay said more detailed automation is still possible through traditional application programming interfaces. In the same implementation session, she built a utility that took PDF and image receipts, categorised them and pushed them into Xero as spent money transactions and bills, with attachments included.
That example suggests the newer conversational connectors may, for now, be more useful for management overviews than for end-to-end accounting workflows. Their immediate value is likely to lie in retrieving headline financial information in a chat interface rather than replacing operational steps in bookkeeping and compliance processes.
"That will probably expand. But right now, if you're hoping to use this to replace the manual Xero export step in your workflow, you're not there yet," said Rodnay.