About
Two developers, one library you can read.
MentorAgent is written and maintained by Emanuele Longo and Antonio Macrì. It's open source: what it does, and what it doesn't, is in the code.
The people
We are the two authors of the library, as listed on NuGet.
-
Emanuele Longo
Author and maintainer of MentorAgent
Software architect with years of work on .NET and enterprise systems. He focuses on architecture, integration and bringing AI into applications in a governed way.
-
Antonio Macrì
Author and maintainer of MentorAgent
Senior .NET software engineer with long experience building applications and libraries. He leads the development of MentorAgent and looks after its quality and releases.
Why MentorAgent exists
Putting an AI agent in front of a demo is easy. Putting it in front of real users, on real data, is not: someone has to decide who may do what, when the agent must ask before changing something, how much it may spend and what leaves the application.
We wanted those decisions to live where the rest of the application's rules already live: in the code, checked the same way on every channel. MentorAgent is the library we built for that, for .NET applications first and for any front end and agent around them.
Your code decides what the agent can do. The model only chooses among the actions you allow.
What we stand for
Six commitments behind the library, the documentation and this site, each with the page that shows it.
-
You stay in control
The agent does only what your code allows, and a person decides on anything that changes data. Governance is part of the design, not an add-on.
-
Your data stays yours
A cloud model or a local one: the library runs where your app runs, and nothing reaches us. No telemetry.
-
Open by choice
Open source under Apache 2.0, built on open standards (MCP, A2A, OpenTelemetry), with no lock-in to us. The agent reaches any front end.
-
Honest about what it does and doesn't
We say when MentorAgent is not a good fit, what you still have to protect and what leaves your application.
-
Quality you can check
Every version is checked for breaking changes, every package lists what's inside it (SBOM), and every build traces back to its commit.
-
Built to last
Semantic versioning, support windows for older versions and response times for security reports.
What's coming
A console for your agents
We're working on a console to manage the MentorAgent agents in your applications. We'll present it here when it's ready, with no dates until then.
Get in touch
Public questions help the next person too: start on GitHub when you can.
-
Questions and ideas
How to do something, what you'd like to see next.
-
Bugs
Something doesn't work as the documentation says.
-
Security
A vulnerability: write privately, not in a public issue. The security page says how.
-
Everything else
Articles, collaborations, or anything that doesn't fit GitHub.