AI · MCP server

Run your insurance platform from the assistant your team already uses

Connect an AI assistant like Claude straight to Openkoda over the open Model Context Protocol. Your team asks questions and gets work done in plain English - against your live book - with every action scoped to their permissions, confirmed before it happens, and on the audit trail.

What it is

Your whole PAS, one conversation away

The Model Context Protocol (MCP) is an open standard for connecting AI assistants to real systems. Openkoda includes an MCP server, so the assistant your team already uses can work directly with your policies, claims, billing and reports - no new screens to learn, no data exports, no custom integration project.

Works in tools you already use

Connect Claude Desktop or any MCP-capable assistant. Your team works where they already are.

Plain English, no training

Ask for what you need in everyday language - there is no query language and no new interface to roll out.

Safe by design

Every request runs with the user’s own permissions, write actions need explicit confirmation, and everything is audited.

What you can ask it to do

Everyday work, in plain language

Real routines your team runs every week - now a sentence instead of a screen tour.

Review your renewal book

“Which policies renew in the next 45 days, with premium and open claims - flag retention risks.”

You get: a ranked shortlist of renewals and the accounts most at risk of leaving.

Triage the referral queue

“Summarise my underwriting referrals - how many are waiting, the oldest, and the top three to action first.”

You get: a clear read on the queue and where to start, without opening ten screens.

Prioritise open claims

“Rank open claims by amount incurred and flag any with large reserves or no recent activity.”

You get: the claims that need attention today, highest-exposure first.

Chase overdue premium

“Show accounts-receivable aging and propose which invoices need a reminder.”

You get: an aged-debt summary and a suggested collection list - you approve before anything sends.

Start new business

“Help me submit new business - list the products, walk me through the form, and create the submission.”

You get: a guided intake that ends with a submission in your system, ready to quote.

Answer a question on the spot

“What's my loss ratio by product this quarter?”

You get: a straight answer from your live data - no analyst queue, no export.

From questions to action

It doesn’t just answer - it does the work

Ask it to renew or endorse a policy, log a first notice of loss, record a claim payment, send a payment reminder, or draft a product change. Anything that changes your data is shown to you as a preview first, and only goes ahead when you confirm - so the assistant saves time without ever acting on its own.

  • Reads your book - policies, claims, billing and reports
  • Acts on request - renew, endorse, file a claim, take a payment, draft a change
  • Every change previewed first; nothing happens until you confirm
Getting started

Live in an afternoon, not a project

Turn it on for your organization

An admin enables the MCP server and chooses read-only or read & write - your call.

Create a secure access token

Generate a token scoped to a user’s exact permissions. It’s shown once, stored safely, and you can revoke it anytime.

Connect your assistant

Add the Openkoda address and token to Claude Desktop or any MCP-capable client - your IT team can do it in minutes.

Safe by design

Ready for real insurance work

The same governance as the rest of the platform - so you can put it in front of the team with confidence.

Runs as a real person

The assistant acts with the connected user’s own role and permissions - it can never see or do more than they can.

You approve every change

Write actions are previewed and require explicit confirmation. A connection can be locked to read-only per organization.

Fully audited

Every request and result is logged, attributed and timestamped - the same audit trail as everything else.

Your data stays put

Requests run inside Openkoda against your own system, returning only what the user may see; tokens are stored hashed and revocable instantly.

The payoff

Why teams turn it on

Faster answers

Numbers and next steps in seconds - not a ticket in the analyst queue.

Less clicking

Skip the screen-by-screen navigation for routine questions and tasks.

Safe for production

Permission-scoped, confirm-gated and audited - built for real insurance decisions.

FAQ

Questions, answered

Do we need developers to use it?
No. Business users work in plain English through an assistant they already know. Setup is a one-time step - enabling the server and connecting a client - that your IT team can complete in minutes.
Which assistants can connect?
Any client that supports the Model Context Protocol, such as Claude Desktop. MCP is an open standard, so the set of compatible tools keeps growing.
Does our data leave our system?
Requests run against your own Openkoda instance and return only what the connected user is allowed to see. Access tokens are stored hashed and can be revoked at any time.
Can it change things on its own?
No. Anything that changes data - a renewal, an endorsement, a payment, a product change - is shown to you as a preview first and only proceeds when you explicitly confirm.
What can it access?
Exactly what the connected user can, and nothing more. Every call runs with that person’s role and permissions, scoped to your organization, and is fully audited - and a connection can be locked to read-only.
Which edition is it in?
The MCP server is part of Openkoda Enterprise. See pricing or talk to us about enabling it.
How is it billed?
The MCP server is included with Enterprise. The assistant you connect can be your own (such as Claude Desktop), or you can use AI features powered by Openkoda AI, our monthly AI subscription.

See Openkoda in your context

Book a live, personalized demo with our product team - tell us your use case and see the platform work with your data. No commitment.