Nance
Nance

Nance + MCP.

Connect anything

Model Context Protocol is the open standard for giving AI agents access to tools and data. Nance is an MCP client: she connects out to the MCP servers you add and allow, tool by tool, so your stack does not end where our integration list does.

Every integration list runs out eventually. The tools that matter most are often the ones nobody builds connectors for: the internal ordering system, the industry-specific ERP, the database that one team swears by. Traditionally that means exports, copy-paste work or an integration project.

MCP changes the economics. It is an open protocol that lets an AI agent like Nance use external systems through a standard interface. Anything with an MCP server in front of it becomes something Nance can read from and act on, with the same permissions, approvals and audit logging as our native integrations. Your finance agent grows with your stack instead of capping it.

Why connect Nance to MCP.

The connection and the colleague in one

She reads MCP, applies your rules and does the actual finance work in your accounting system.

Nothing posts outside your rules

You decide what runs automatically and what waits for approval. Every action logged, with the why attached.

One agent across the stack

Payments, commerce, CRM, spend and inboxes: one set of rules and a full audit trail between them.

Up and running in three steps.

  1. Connect

    Through the official MCP API, with scoped access. Nothing to install and nothing to migrate.

  2. Set the rules

    What she books on her own, what waits for you, who approves. Start anything as a Test Run: a read-only preview first.

  3. She works daily

    What arrives in MCP is read, matched and booked in your accounting system. Exceptions come to you with context.

What Nance automates with MCP

Internal systems

Your own ERP, ordering system or data warehouse can be exposed to Nance through an MCP server, and becomes part of the same workflows as your accounting system.

Niche and legacy tools

The industry-specific package nobody integrates with? If it has an API, a thin MCP server in front of it is usually all Nance needs.

Read and act across systems

Nance can read context from one system and act in another: check a contract system before paying, verify stock before invoicing.

Governed access

Every MCP connection is scoped: which tools, which actions, read or write. Approval gates apply to MCP actions exactly as they do everywhere else.

Full audit trail

Every call Nance makes over MCP is logged with its context and outcome, so connecting more systems never means losing oversight.

No throwaway plumbing

MCP is an open standard, not our proprietary connector format. The servers you build or buy keep working with the rest of your AI tooling.

Access and security

  • No bank connection, no bank credentials, ever
  • Her own user account with scoped permissions
  • Every action logged, visible and attributable
  • Test Run: read-only preview before anything changes
  • Hosted on Google Cloud in the EU
  • Your data never trains AI models

Frequently asked questions

Connect MCP to Exact Online, Twinfield, Moneybird, Xero and QuickBooks

Most MCP connectors are data pipes: they copy fields on a schedule and stop there. Nance is the connection and the colleague in one. It reads MCP, applies your rules and does the actual finance work in your accounting package, with approvals where you want them and every action logged.

Security is part of the job.

Nance works inside your financial systems, so we treat protecting them as core product work. The live details are public in our trust center.

Encrypted everywhere

AES-256 at rest, TLS in transit.

GDPR compliant

a standard DPA with every plan.

ISO 27001

certification in finalization.

Hosted in the EU

on Google Cloud, by default for everyone.

Never trains AI models

your data stays yours alone.

Payments stay at your bank

Nance prepares the batch, you release it.

Hire Nance.

Connected to MCP in an afternoon, and we walk you through it.