Skip to main content

The Aigentic vision

A demo agent is easy to build. A production agent needs more: a way to see what it did, check that it still works after a change, and improve it without guessing. Aigentic is built around that loop.

  1. 01

    Build

    Define the agent in Kotlin: typed input and output, a task, tools and a model.

    agent<I, O> { }
  2. 02

    Run

    Run it inside your application on the JVM or Node.js, with the provider you choose.

    agent.start(input)
  3. 03

    Record

    Each run goes to the Aigentic platform with its messages, tool calls and token usage.

    platform { }
  4. 04

    Tag

    Review runs on the platform and tag the ones that went right.

    Aigentic platform
  5. 05

    Replay

    Replay tagged runs as regression tests. Tool calls return the recorded results, so nothing touches your real systems.

    regressionTest { }
  6. 06

    Improve

    Change a prompt, a tool or the model, and ship once the tests pass. Then the loop starts again.

    back to Build

Build and Run need only the core library and a provider. Record, Tag and Replay use the platform module and the Aigentic platform. Flock hosts the platform and gives access on request. See the testing example for a regression test in code.