Atlas is a terminal-native AI coding agent for Quarkus, the Java framework where build-time augmentation and native-image constraints shape what code you are allowed to write. In 2026 you run atlas in a project with a pom.xml that pulls in quarkus-bom, and Atlas reads your CDI beans, JAX-RS resources, and the properties in application.properties.
Quarkus, build-time augmentation, and why Atlas fits
Quarkus performs build-time augmentation, which is why native-image constraints shape what code you are allowed to write in 2026. Atlas searches code with hybrid semantic and keyword retrieval fused by reciprocal rank fusion, so a reflective call site is found by its literal text and by its meaning at once.
Quarkus moves work that other Java frameworks do at startup into the Maven build step. That is why a Quarkus service starts fast, and also why reflection-heavy Java that compiles cleanly and runs happily on the JVM can still collapse when the native-image build closes the world. Hybrid retrieval matters here because the thing you are hunting is often a literal, a Class.forName string or an annotation name, sitting inside code whose purpose you can only describe semantically. Reciprocal rank fusion combines both rankings, so the reflective bean surfaces.
Reading CDI beans, JAX-RS resources, and application.properties
Point Atlas at a project with a pom.xml that pulls in quarkus-bom, and it reads your CDI beans, JAX-RS resources, and the properties in application.properties. In Quarkus, 1 behavior is often split in half, an annotation on the class and a key in a properties file.
Reading only one half of a Quarkus configuration produces confident wrong answers. Atlas maps the CDI bean scopes, the JAX-RS resource classes with their paths and their producers, and the config keys those classes pull in through @ConfigProperty, then holds the annotation side and the application.properties side together. Ask why an endpoint 404s under one profile and Atlas can compare the JAX-RS path against the profile-specific property that disables it, instead of reading the resource class alone and telling you the route looks fine.
Replacing runtime reflection so the class survives native-image
Atlas replaces runtime reflection with a build-time-friendly pattern so the class survives native-image, and it shows the diff before writing. Quarkus closes the world at build time, so a Java class reached only through reflection is dropped from the binary, and in 2026 the failure still appears long after the JVM build passed.
Native-image cannot see a class that is only ever reached reflectively, so it eliminates it, and the Quarkus build fails or the binary breaks at runtime on a path nobody tested. Atlas locates the reflective access, then proposes a build-time-friendly pattern that keeps the behavior while removing the runtime lookup that the native build cannot follow. Atlas computes a unified diff for every file edit and surfaces it for approval before writing, so you compare the rewritten bean against the original rather than accepting a Java rewrite on trust.
@QuarkusTest, RestAssured, and ./mvnw test
Atlas adds @QuarkusTest cases with RestAssured and runs ./mvnw test behind a permission prompt, gated by 3 rule outcomes, allow, ask, and deny. A @QuarkusTest boots the real Quarkus application, so RestAssured drives the JAX-RS resource, its CDI injections, and the application.properties config as one running service.
Testing a Quarkus resource in isolation proves very little, because the interesting failures come from CDI wiring and configuration rather than from method bodies. A @QuarkusTest starts the application and RestAssured issues real HTTP requests against it, which catches an unsatisfied bean or a missing property that a mocked unit test would sail past. Every Atlas tool call is permission-gated against allow, ask, and deny rules before it runs, so ./mvnw test executes only after you approve the Maven command.
Spotless, commits, and rollback
Atlas surfaces the diff for review, then applies Spotless before it commits. Quarkus projects are Maven projects with enforced formatting, and Atlas snapshots file changes as git patches, so in 2026 a native-image refactor across several CDI beans is rolled back as one unit rather than manually repaired.
Atlas closes out a Quarkus change the way a Java reviewer expects. Spotless runs over the modified sources so formatting never shows up as review noise, and the change lands as a commit Atlas can stage on your behalf after reading git branches, status, and diffs. Because every edit was snapshotted as a git patch, a reflection rewrite that passes ./mvnw test but breaks the native build is one rollback away, not an afternoon of undoing hand edits across a Maven module.
Getting started
- 01Run atlas in a project with a pom.xml that pulls in quarkus-bom
- 02Let Atlas read your CDI beans, JAX-RS resources, and the properties in application.properties
- 03Ask Atlas to replace runtime reflection with a build-time-friendly pattern so the class survives native-image
- 04Let Atlas add @QuarkusTest cases with RestAssured and run ./mvnw test behind a permission prompt
- 05Review the diff, then let Atlas apply Spotless before it commits
Frequently asked questions
- how to use an AI coding agent with quarkus
- Run atlas in a project with a pom.xml that pulls in quarkus-bom. Atlas reads your CDI beans, JAX-RS resources, and the properties in application.properties, then proposes edits as unified diffs you approve.
- why does my quarkus native-image build fail when the jvm build works
- Native-image drops classes reached only through reflection. Ask Atlas to replace runtime reflection with a build-time-friendly pattern so the class survives native-image.
- can atlas write quarkustest tests with restassured
- Yes. Atlas adds @QuarkusTest cases with RestAssured, which boot the real application and drive the JAX-RS resource over HTTP, then runs ./mvnw test behind a permission prompt.
- does atlas understand quarkus cdi bean scopes
- Yes. Atlas maps your CDI beans and their scopes alongside the JAX-RS resources that inject them, so an unsatisfied bean is traced to the producer that should have supplied it.
- how do i format java code after an AI agent edits it
- Review the diff, then let Atlas apply Spotless before it commits, so formatting never appears as noise in the Maven project's code review.
- can atlas index a private java codebase locally
- Yes. Atlas can build its code index with local Ollama embeddings, keeping code off third-party servers, which suits enterprise Quarkus services.
- can i roll back an atlas change to a quarkus bean
- Yes. Atlas snapshots file changes as git patches so edits can be diffed and rolled back, and it can stage and create commits on your behalf.
Try Atlas in your terminal
The terminal-native AI coding agent. Free core, single binary.
Install AtlasRelated guides
Trace a runtime bug from a stack trace in Quarkus with Atlas in 2026
Pinpoint and fix Quarkus runtime bugs from production stack traces using Atlas, the terminal-native AI agent. Leverage Maven, JUnit 5, and Spotless for rapid resolution.
Refactor a legacy module in Quarkus with Atlas in 2026
Streamline Quarkus module refactoring in 2026 with Atlas. Safely restructure old code, maintain behavior, and prevent breaking callers using Maven, JUnit 5, and Spotless.
Extract a Shared Helper from Duplicated Quarkus Code with Atlas in 2026
Streamline your Quarkus applications in 2026 by extracting duplicated logic into shared, tested helpers using Atlas. Find semantic duplicates, refactor with confidence, and ensure code quality with Maven and JUnit 5.
Upgrade a Quarkus Dependency and Fix Breakage with Atlas in 2026
In 2026, Atlas helps Quarkus developers upgrade major dependencies, fixing compile and test failures. It drives Maven, reads compiler output, and repairs code with LSP, ensuring JUnit 5 and Spotless compliance.
Onboard to an Unfamiliar Quarkus Codebase with Atlas in 2026
Quickly build a working mental model of any Quarkus codebase in 2026 with Atlas. Leverage semantic search, AST indexing, and direct interaction with Maven, JUnit 5 (@QuarkusTest), and Spotless to understand unfamiliar
Diagnose a hanging or long-running command in Quarkus with Atlas in 2026
In 2026, Quarkus developers use Atlas to diagnose hanging or slow Maven builds and scripts. Quickly identify if a command is blocked on input or genuinely slow, and get unstuck.
Debug a Single Failing Test in Quarkus with Atlas in 2026
Pinpoint and fix failing Quarkus tests efficiently with Atlas. Leverage JUnit 5, Maven, and Spotless for a streamlined debugging workflow in 2026.
Write Unit Tests for Untested Quarkus Code with Atlas in 2026
In 2026, use Atlas to write robust unit tests for untested Quarkus modules. Atlas understands your existing JUnit 5 (@QuarkusTest) and Maven conventions, ensuring new tests integrate direct.