Digits has launched Write MCP, an update that lets customers’ own AI agents make changes in the general ledger instead of only reading from it. The feature brings AI agents into everyday accounting work while keeping a human approval step around records that have already been closed.
The key protection is Locked Periods, a new part of Digits’ AI-native accounting platform. Once a period is locked, no person or agent can change it silently. Any proposed edit goes into a review queue, where an admin must approve or reject it before the change reaches the books.
That distinction matters because writing to an accounting ledger is different from asking an AI agent to summarize information. An agent may be able to complete work directly in Digits, but closed periods still remain protected from unapproved changes.
How Write MCP Handles Ledger Changes
Write MCP allows agents to create, update, and delete records in four core areas: Transactions, Bills, Invoices, and Chart of accounts. In open periods, work flows as usual. When an action touches a locked period, the change waits for review instead of posting immediately.
Agents also act under the signed-in user’s permissions. They can only reach the records and areas that user can access, which keeps the agent’s capabilities tied to the account already in use. Actions in locked periods still require human review and approval, even when the signed-in user has permission to work with those records.
Digits designed Write MCP to work with Claude, ChatGPT, Cursor, and any MCP-compatible tool. The server is free, and Write MCP is available at no cost to Digits customers as of October 13, 2026.
In practical terms, an accountant can tell an agent to find and move Stripe fees or review and book recurring expenses. The agent then makes those changes directly in Digits, unless the work touches a locked period, in which case it sends the proposed edit to the review queue.
Why Locked Periods Matter for AI Accounting
Closing a period gives an accounting firm a clear point of control. Digits’ approach keeps that control in place while allowing agents to handle work in periods that remain open. The result is a setup where an agent can do accounting work without giving it silent access to records that a firm has already finalized.
Jeff Seibert, CEO of Digits, described the challenge this way: “Giving an agent write access to the ledger is easy. Doing it safely is the hard part, and that’s what accountants care about most.”
He also explained why Digits built the protection before adding write access: “We built Locked Periods first so we could say yes to what our customers kept asking for. Now your agent can do the accounting, and the books you’ve closed stay closed.”
That order of operations puts the review process at the center of the feature. An agent can propose or complete work, but changes to locked periods remain visible to an admin and cannot post until a person approves them.
An Early Test With Arclow
Arclow, an AI-native accounting practice built on Digits, was among the first firms to use Digits MCP on client engagements. One Arclow accountant used Claude on Digits MCP to complete work in about one hour and 40 minutes instead of seven hours.
Kangsan Wyi, Founder & CEO of Arclow, described the result: “My staff accountant ran two clients’ Q3 closes through Claude on the Digits MCP. Work that takes about seven hours by hand was done in an hour and forty minutes, roughly four times faster. He finished in the morning and went to lunch with peace of mind.”
The example shows the main promise of Write MCP: agents can take on direct accounting tasks, while the firm keeps a formal checkpoint for work involving locked books. That combination separates routine execution from final control rather than treating every ledger change the same way.
Digits was launched in 2025 and now manages $850 billion in financial activity across hundreds of accounting firms and thousands of businesses. With Write MCP, the company is extending its free MCP server from read access to controlled ledger changes, giving customers a way to use AI agents inside accounting workflows without removing human approval from closed periods.
Based on
