In the last three posts, we walked through Sage Intacct’s five AI agents and what to do with them. That’s one half of the Sage Intacct AI story. The other half is newer and, in some ways, more interesting. Sage now gives organizations a governed way to connect approved external AI tools to Sage Intacct data.
It’s called the Sage Intacct AI Gateway. Underneath it sits something called an MCP Server, which sounds like something that should live in a locked closet with blinking lights. Thankfully, the basic idea is simpler than the name. If Sage Copilot is for finance users working inside Sage Intacct, AI Gateway is for organizations that want approved external AI tools to query Sage Intacct data through a governed connection.
The plumbing
MCP stands for Model Context Protocol. It’s an open standard that lets AI tools connect to business systems through a consistent, governed interface.
The Sage Intacct AI Gateway uses MCP plus Sage Intacct REST APIs to connect approved AI applications to Intacct data. That could mean an external LLM, an internal AI tool, or a partner-built application that needs financial context from Sage Intacct.
Sage’s own metaphor here is actually pretty good: think of it as a USB port for AI. Plug in a compatible, approved tool, and it has a governed way to access the data it is allowed to see.
A few specifics worth knowing:
- Sage introduced AI Gateway with Sage Intacct 2026 R2, with availability in multiple regions, including the US, UK, Canada, Australia, South Africa, and Singapore.
- MCP Server access is read-only at launch and is designed for AI discovery, reporting, assistant, and analysis workflows.
- REST API access can support developer-built integrations that read and write Sage Intacct data, depending on permissions and API capabilities.
- Role-based permissions carry through, so connected tools should only access what the authenticated user or integration is allowed to access.
- Two Sage licenses are required: a Web Services Developer License and an AI Gateway Developer License.
- API transactions count. Every query an AI tool makes counts toward Sage Intacct’s standard 100,000-transaction monthly limit. The MCP Server doesn’t filter or rate-limit what the AI asks.
Why this matters
The first three blogs focused on choosing among Sage’s own AI agents. The AI Gateway opens up a second decision, which is whether Sage’s tools should be your AI interface, or whether you’d rather point your own AI at Sage’s data.
The good news is you can have both. A small or mid-market finance team without its own AI infrastructure will likely benefit more from Sage Copilot and the agents we discussed earlier in the series. A larger or more technically advanced organization may already have an LLM strategy and want Sage Intacct data integrated. You can now do either, or both, with a more standardized connection path than building everything from scratch.
What this looks like in practice
People are already running this kind of analysis (primarily with QuickBooks) using the same underlying technology. A few ideas we’ve heard about include:
- Natural-language reporting. Instead of clicking through filters and exporting to Excel, you ask, “what did we spend on professional services in Q2 by entity?” and get the answer.
- AR aging and collections. “Which customers are over 60 days, what’s the total exposure, and what’s their payment history?” — answered, ranked, with the follow-up emails drafted if you want them.
- Anomaly and exception review. “Any expenses this month that are unusually large for their account?” or “Which customers’ aging is worse this month than last?” The kind of work that used to mean pulling three reports and reading them side by side.
- Cross-system analysis. Pair Sage Intacct data with your CRM, billing platform, or marketing spend, and ask questions that span them. “How does actual revenue compare to what sales marked closed?”
Keep in mind, though, that because the Sage Intacct MCP Server is read-only at launch, an AI tool connected through it can’t post entries, modify records, or send anything on its own. It can pull, analyze, and draft (the variance commentary, the reminder email, the board narrative) all from your live data, all for your review before it goes anywhere.
Before you plug anything in
There are a few things worth thinking about before you let an external AI start asking questions of your financial data.
Data quality matters more, not less. We made this point about the Finance Intelligence Agent in the last post. The AI Gateway raises the stakes. Now any AI tool can query your data, and dirty data produces confidently wrong answers across more interfaces.
Read-only at launch is doing important work. It means an AI tool, no matter how confident it sounds, cannot post a journal entry or change a vendor master. That’s the right starting posture for finance. Don’t be in a hurry to expand beyond it.
BYO-AI policy is a real conversation now. Who can connect an AI tool to Intacct? Which tools are approved? What data can be queried? Can results be copied into other systems? Who reviews prompts, outputs, and access logs? Many companies have not written this down. They should do so before the first curious power user connects something “just to test it.”
The 100K API limit isn’t new, but the AI joining the queue is. Every question your connected AI asks counts toward your monthly Sage Intacct API allowance, which your existing integrations are already using. An AI exploring your data can consume transactions differently from integrations, and the MCP Server doesn’t control the calls on your behalf. Keep an eye on usage and plan accordingly.
One last thing
The bigger question AI Gateway opens up is not just which AI tools to connect. It’s what access to give them, who approves that access, and how your organization governs usage once financial data can be queried outside the normal Intacct screens.
Most mid-market finance teams have not had that conversation yet, and they should.
If yours is starting to, we’re around. We can help you think through where AI Gateway fits, what needs to be cleaned up first, and how to keep the exciting new plumbing from turning into exciting new problems.
