Custom AI for your business
AI with controlled access to business data.
Less time on reports.
More time for your business.
We design custom MCP connections to accounting and other business systems. Ask questions about your business in everyday language; first, we define which data and actions the AI may access.
Let’s talkA custom project with a separate quote. This demonstration uses fictional data, not a ready-made accounting connection. Scope, permissions and security requirements are agreed with you before systems are connected.
Choose a question and explore the answer. No real systems are connected here.
Today’s sales are 15.2% higher than at the same time last Wednesday.
Where does the answer come from?
Fictional example: completed point-of-sale and online sales on 4 November 2026 up to 14:32. Figures exclude VAT and are net of refunds. The comparison covers the same period on 28 October. An online order also recorded at the till is counted only once.
The sand-coloured coffee cup sold the most units today.
Coffee cup · sand
24 sold today
| Product | Units |
|---|---|
| Coffee cup · sand | 24 |
| Travel flask · black | 19 |
| Guest towel · natural | 12 |
Fictional point-of-sale and online sales · 4 Nov, up to 14:32
Two products are below the minimum stock levels set in this demonstration.
| Product | Available | Minimum |
|---|---|---|
| Travel flask · black | 4 | 10 |
| Guest towel · natural | 6 | 12 |
Available stock accounts for items reserved in orders. The next step would be to check incoming deliveries before placing a purchase order.
Where does the answer come from?
Fictional stock levels and thresholds from an inventory system, alongside reservations from an order system. Status at 14:32 on 4 November 2026. This demonstration does not place any purchase orders.
Let’s put the coffee cup on sale until Sunday. You review and confirm the change first.
Coffee cup · sand
SKU KB-104 · prices including VATValid 4–8 November 2026. The previous price resumes when the offer ends.
Choose the online shop, point of sale, or both.
The demonstration offer is confirmed.
This shows the confirmation process. No prices have changed in real systems.
Access defined in advance
The right information.
The right permissions.
Gonora MCP can provide the control layer between business systems and AI. For each project, we define which queries are allowed and which information may appear in the results.
- 01
Accounting and operations
Your systems and data. Access is agreed with the people responsible in your business.
- 02
Gonora MCP
Scoped permissions, filtered responses and action logging as defined in the implementation.
- 03
Approved AI service
ChatGPT, Claude or another chosen tool receives the results returned by the connection.
Less access
Select only the queries needed and separate reading from editing.
Less data
Design responses using summaries and selected fields instead of sending entire datasets.
Revocable access
Define how to stop a connection and revoke a user’s or tool’s access.
Traceable actions
Define what is logged, who may view the log and how long it is kept.
Results are passed to the AI provider. MCP is an access interface, not a sealed data boundary. Data in a tool response may go to ChatGPT or Claude, where it is processed under the service plan, settings and agreement your business uses.
Explore MCP and the setup guideAnswers from your data.
Actions in your systems.
We start with one useful question, for example about sales or cash flow. Then we assess data quality, connection options and the permissions needed to answer it.
Answers you can trace
See the data source, reporting period and last update. We define what figures such as sales, gross margin and available stock mean in your business.
Your permissions govern access
The implementation defines role-based access, preferably starting with read-only permission. Actions such as price changes require separate permissions and confirmation steps, as shown in the demonstration above.
Connections for your systems
We assess API access, agreements and update frequency for your accounting, point-of-sale or ecommerce systems before agreeing the scope. Each connection is a separate development and testing project; no specific accounting provider is claimed to be connected in advance.
Where does business data go?
When AI calls an MCP tool, it receives the tool’s result, which may contain business data. We design scoped queries, filtering and summaries to reduce the amount of data. Your business must also choose an appropriate AI service, agreements, retention settings and access permissions.
Is the data used to train AI?
That depends on the provider, service plan and your choices. OpenAI says business-service and API data is not used for training by default. Anthropic describes a comparable default for its commercial services, with exceptions including data explicitly supplied as feedback. This is not a blanket statement about personal accounts. See OpenAI’s business data policy and Anthropic’s explanation of training and data.
Is this the same MCP that costs 1,500 ISK?
No. MCP access to a Gonora website costs 1,500 ISK/month excluding VAT (about €10.98) and is intended for editing your own website content. Accounting and business-system connections are custom projects with separate quotes. An AI subscription, model usage, external-system access and development of those connections are not included in the website MCP add-on.
How do we agree security requirements?
Before the project starts, we review data types, permissions, retention and agreements with you. MCP alone guarantees neither secure processing nor compliance with privacy requirements. The implementation, choice of providers and your business procedures must meet the requirements applicable to your data.
Which question would you like to answer in seconds?
