Financial services firms across capital markets, banking, insurance and payments run on code they’ve built over years: pricing and risk libraries, credit decisioning engines, claims and actuarial models, fraud and payments rules. Much of it is written in Java, and none of it lives in a public package registry.
AlphaAgent was already the only agent platform that let you build your own execution environment, bringing your own Python packages in a Docker image, today we’re taking it further.
AlphaAgent agents can now call your own Java libraries, in execution environments you build, running on AWS Lambda inside your own AWS account.
What this means for customers:
Your libraries, not a rewrite. Agents call the pricing, credit, actuarial and fraud logic your teams already trust, straight from Python through JPype. The answers come from your code, not a model’s approximation of it.
You define the runtime. Start from the AlphaAgent Python and Java image: Java 21, JPype, the full Python data stack and the document toolkit are already in it. Add your JARs and internal Python packages, push to your own registry, and pick it in Studio.
Internal code stays internal. Everything ships inside the image and is pinned by digest. Nothing is fetched at run time, and your libraries never leave your AWS account.
Every run is sandboxed on AWS Lambda. One function per environment, a fresh working directory for every run, and short-lived credentials scoped to one user’s files. The agent’s code itself holds no AWS credentials.
Build once, use across the firm. Share an environment through AlphaAgent Libraries so every team runs the same pinned image. Studio upgrades deliver the execution handler at run time, so your custom images never need rebuilding for them.
For engineering teams, it’s a short Dockerfile: start from the AlphaAgent Python and Java image, copy in your JARs, install your requirements, push. Studio provisions it as a Lambda function and checks it before any agent can use it.
Docs: https://docs.alphaagent.prometheusrl.com/platform-internals/how-agents-run-code#build-a-custom-environment-image-on-the-python-and-java-image