AI Is Coming to midPoint, and It Is Solving the Right Problem First
midPilot uses generative AI to accelerate application onboarding in midPoint — the part of IGA deployment that consumes the most engineering time. Here is what it does and why it matters.
I’ve written three midPoint connectors from scratch over the past couple of years: one for AWS, one for GCP, and one for Convenia HR. All three are open source and free for the community to use. I mention this not to promote them but because building those connectors gave me a very direct understanding of what makes application onboarding into midPoint slow, and why what Evolveum is doing with midPilot matters.
What the Work Actually Looks Like
The process of onboarding a new application into an IGA system follows the same general shape every time. You read whatever documentation the source system provides, figure out how it represents users, groups, and organizational units, then map that to midPoint’s schema. You write the connector code to handle each operation: authentication, object discovery, provisioning, and deprovisioning. Then you work through the correlation logic to decide how accounts on the source system relate to identities already in midPoint.
Each of these steps involves judgment calls, and many of them require going back and forth between the source system’s behavior and what you’ve configured until it works correctly. For well-documented, widely adopted systems like AWS the process is tractable. For an internal HR system or a SaaS tool with a thin API, it can take weeks.
What midPilot Does
midPilot is Evolveum’s answer to this. Rather than leaving the engineer to do all of that discovery and translation manually, midPilot uses generative AI to generate the connector code, suggest attribute mappings, and recommend the correlation rules for linking accounts and roles. The engineer reviews and confirms at each step — which is the right design for production identity infrastructure where you don’t want autonomous decisions you haven’t seen. The whole interaction is iterative and human in the loop throughout.
The project ran from April 2025 through June 2026, funded through the European Union’s Recovery and Resilience Plan of the Slovak Republic. A preview landed with midPoint 4.11 in June, and the full release is coming this October. The complete feature set, including the Integration Catalog where the community can share connectors, mapping configurations, and correlation rules, arrives with midPoint 4.12 LTS in 2027. You can track the full roadmap at docs.evolveum.com/midpoint/roadmap.
The October release is worth watching closely if you are working in IGA. midPilot directly addresses the part of IGA deployment that consumes the most engineering time, and the Integration Catalog has the potential to shift how the community shares and reuses connector work.
We Can Help
Atricore is an Evolveum partner. If you are evaluating midPoint or planning a new integration, our IGA leader would be happy to connect.