Claude Agent in IntelliJ IDEA Public Beta: How Java Development Is Changing

Claude Agent enters IntelliJ IDEA public beta

On June 22 GitHub announced that Claude can now act as an Agent Provider in GitHub Copilot for JetBrains IDEs, allowing developers to install the Claude Code CLI, configure its path in IntelliJ IDEA settings, and select Claude from the Copilot Chat Agent picker.

Why Java developers should care

Java projects typically consist of many Maven modules, layered packages (Controller, Service, Domain, Repository, Mapper, DTO, configuration, tests), and complex runtime dependencies such as Redis, Feign, MQ, and transaction interceptors. Understanding the full call chain—e.g.,

OrderController → OrderApplicationService → OrderDomainService → OrderRepository → OrderMapper.xml → MySQL

—is essential for troubleshooting.

AI coding shifts from code completion to task execution

Previously AI assistants only offered code completion: generating method bodies, DTO fields, or Mockito snippets. The new Agent mode lets developers hand over an entire engineering task, such as analyzing a startup failure, locating the root cause, optionally modifying code, running tests, and presenting a diff.

Real Spring Boot failure example

A recent change introduced a risk‑check feature, causing the service to fail with the error “Parameter 1 of constructor in OrderService required a bean of type 'com.example.risk.RiskClient' that could not be found.” Traditional debugging requires inspecting the stack trace, searching for RiskClient, checking Feign annotations, Maven dependencies, profiles, and conditional beans.

With the Agent, the developer can issue a structured request:

请只读分析当前 Spring Boot 启动失败问题，不要先修改代码。
错误提示为 RiskClient Bean 不存在，请按下面的顺序检查：
1. RiskClient 的声明方式和所在模块
2. order-service 是否引入对应依赖
3. 启动类是否启用了正确的 Feign 扫描
4. 是否有 Profile 或 Conditional 注解阻止 Bean 创建
5. 最近 Git Diff 是否改变了相关配置
最后输出：
- 最可能的根因
- 支持该结论的代码位置
- 最小修改方案
- 修改后需要运行的验证命令

The Agent reads logs, pom.xml, Spring configuration, and recent Git changes, then returns a concise root‑cause analysis, code locations, a minimal fix, and verification commands.

IDE will not be replaced, it becomes more critical

Even though the Agent can modify dozens of files, developers still need the IDE to review changes, run inspections for potential null pointers, examine Maven dependency graphs, and debug runtime behavior. IntelliJ IDEA’s role shifts from “helping write code faster” to “helping review the Agent’s work.”

Team‑level custom Agents

Organizations can publish unified custom Agents that appear automatically for all members. Examples include a “Spring Boot Code Review Agent” that checks interface contracts, transaction boundaries, MQ idempotency, Redis consistency, and mapper SQL risks, or a “Startup Fault Analysis Agent” that enforces a read‑first, modify‑later workflow.

Risk of bypass‑permissions mode

The current implementation runs with bypass permissions, meaning file edits and tool invocations are auto‑approved. For production projects this poses a risk, so developers should adopt safeguards:

Ensure a clean Git working tree before starting an Agent task ( git status).

Avoid running the Agent in environments containing production credentials.

Prefer isolated branches or worktrees for complex tasks.

After each task, review the diff, run tests, and verify the build manually.

Getting started

Install the latest GitHub Copilot JetBrains plugin and the Claude Code CLI. Verify the CLI path with which claude (or where claude on Windows), then configure the path in Settings → Tools → GitHub Copilot → Chat and select Claude from the Agent picker.

Recommended initial tasks

Fault analysis: Spring Boot startup failures, bean injection errors, Maven conflicts, test failures.

Impact analysis: Determine which modules or services are affected by changes to an enum, DTO, or database field.

Test generation: Ask the Agent to list test scenarios and generate JUnit 5 + Mockito tests for a specific service.

Effective task template

任务背景：order-service 在新增风险校验后启动失败。
当前目标：定位 RiskClient Bean 无法创建的根因，并给出最小修改方案。
分析范围：
- order-service
- risk-client 模块
- 启动类和 Maven 依赖
- 当前 Profile 配置
- 最近一次 Git Diff
禁止事项：
- 不要修改 common-core
- 不要修改 application-prod.yml
- 不要执行数据库写操作
- 分析完成前不要修改代码
输出要求：
- 根因判断
- 相关文件和代码位置
- 最小修改方案
- 可能影响范围
- 修改后验证命令

From personal assistant to team engineering role

Agents will evolve into multiple team‑defined roles—code review, fault analysis, test generation, dependency upgrade, security scanning—while developers focus on defining tasks, correcting direction, validating results, and making final decisions.

Conclusion

Claude Agent’s public beta turns IntelliJ IDEA into a command‑center for AI‑driven engineering. It does not eliminate coding but shifts developer effort toward system understanding, task orchestration, result verification, and risk management.