Connect AFAS Profit to an AI agent

What is an AI agent connected to AFAS Profit?

An AI agent connected to AFAS Profit answers questions in plain language with data it reads from AFAS through the official API. Where you allow it, it prepares an action, such as a new dossier item. It has only the rights you give it, and a colleague approves every change it proposes.

What does AFAS's own AI, Jonas, do in Profit 8?

Jonas is the AI assistant in AFAS Profit 8, released in mid July 2026. It fills workflow fields, reads documents, transcribes speech and translates, inside Profit, InSite and Pocket. As of September 2026 it costs nothing extra, under a fair use policy. It does not serve customers on your website or connect other systems.

• Fills workflow fields such as subject, urgency, priority and summary, and marks what it filled.
• Analyses purchase quotes and applications, and runs custom assistants on your own documents.
• Helps with recruitment matching, tax questions and receipts in Pocket. Switched on per environment and function.

Is there an official AFAS MCP server or certified AI integration?

As of September 2026 AFAS has no official MCP server; it is investigating one. In June 2026 AFAS stated that no AI integration is certified yet, and it calls uncertified MCP servers and AI agents risky. BeeManaged is not certified either and does not present itself as an AFAS solution.

AFAS advises using certified solutions only. None exist yet, so that choice is yours. This is how we apply the rest of its advice:

• Minimal token rights, read-only by default.
• A dedicated system user per app connector, so you can trace and block the agent.
• OAuth for every new connection.
• Processing in the EU; your data is not used for model training.
• A standard data processing agreement, and your AFAS administrator switches the connection on.

• What an MCP server is and when you need one

Jonas, a self-service connector or a managed agent: which route fits?

For work that stays inside AFAS, Jonas is the obvious choice. When one person wants to ask their own administration questions from a chat tool, a self-service MCP connector will do. Choose a managed agent when AFAS has to work with customer contact or other systems, and when write actions need approval, monitoring and an owner.

Three routes to AI with AFAS Profit (as of September 2026)

What does an AI agent do with AFAS in a manufacturing company?

Mostly it saves searching and retyping. The agent answers customer and order questions, shows project leads their hours, turns incoming requests into dossier items or tasks after approval, reacts to workflow tasks and combines AFAS with conversation data for management.

1. 01

   Customer and order data for sales office and service

   "Which orders does this customer have open?" The agent reads the customer and order GetConnectors and answers during the call.

2. 02

   Projects and hours for project leads

   "Who booked hours on this project this week?" The agent returns a short overview from a projects and hours GetConnector.

3. 03

   Incoming requests as a dossier item or task

   A request arrives by website chat or email. The agent collects missing details and prepares a dossier item or task, written through an UpdateConnector after approval.

4. 04

   Management information across processes

   "How many service requests came from customers with an open order?" The agent combines conversation data with AFAS data.

5. 05

   Reacting to workflow tasks with webhooks

   Since Profit 8, AFAS sends webhooks when dossier items, reactions or workflow tasks change. The agent then fetches the details and prepares the next step.

Power BI with AFAS or an AI agent: what is the difference?

For fixed dashboards, use the official AFAS BI models in Power BI. An agent answers the questions nobody built a dashboard for, in plain language, with data fetched when you ask and combined with sources outside AFAS. The two work side by side; the agent does not replace Power BI.

As of September 2026 the BI models use a BI OData connector with pre-calculated data and template dashboards. They require AFAS Online and Profit 7 or later, refresh at most twice a day, and sharing needs Power BI Pro licences. GetConnectors remain available for custom BI.

• Internal AI agent: ask your own systems in plain language

AFAS in app.beemanaged.com

• The internal agent, where employees ask AFAS and other connected systems questions in plain language.
• Process dashboards that combine conversation data with AFAS data.
• An approval queue for write actions.
• An audit log per question and action.
• Conversation monitoring and a monthly report.

• How the internal AI agent works

How does the AFAS API connection work technically?

The agent uses the AFAS Profit REST API through its own app connector. It is read-only by default: it reads through GetConnectors and writes only through UpdateConnectors after approval. Authentication uses OAuth with separate access per environment, and the agent and its data run in the EU.

• GetConnectors read a data definition set up in Profit; UpdateConnectors insert, update or delete with validation. Special connectors cover dossier items (Subject), reports, images, files and version.
• One app connector per application, linked to a user group, with a dedicated system user. IP restriction and mTLS are optional.
• A test token does not work on production. Endpoint: https://{participant}.rest.afas.online/ProfitRestServices.
• Limits: 100 rows per call by default, 5 parallel calls, 75 MB per call.
• OAuth is the default for new integrations. Classic tokens expire on 1 May 2027; app connectors using one stop working after 31 August 2027.

• MCP compared with a direct API connection

Four steps to an AFAS agent

1. 01

   Choose the process

   One process with a clear question, such as order questions from the sales office.

2. 02

   Set up access

   Your AFAS administrator creates the app connector, system user and GetConnectors.

3. 03

   Build and test

   We build and test in the AFAS test environment, with your team.

4. 04

   Go live and manage

   Writes go through the approval queue; we monitor, adjust and report monthly.

• How BeeManaged works, step by step

What does an AFAS AI agent cost?

An internal agent on AFAS is custom work, priced in consultation. Customer-facing agents that use AFAS data start from €0.35 per conversation. Setup is quoted up front, with development at €175 per hour. Monitoring and the dashboard cost €250 per month, and a consultation package costs €1,750.

• BeeManaged pricing

When do you not need BeeManaged for AFAS?

If the work stays inside AFAS, Jonas is included and may be enough. If one person only wants to ask their own AFAS administration questions, a self-service connector is cheaper, provided you accept it is uncertified. For fixed reports, the AFAS BI models in Power BI do the job.

• Also on Exact Online? Connect it to an AI agent