The Agent Client Protocol, or ACP, is now available in WebStorm, giving developers a standardized way to use different AI agents within a JavaScript and TypeScript development environment, without having to leave the integrated development environment or commit to a single provider.
JetBrains explains that teams that already have subscriptions to Anthropic, OpenAI, or Google can use those services directly in WebStorm without an additional JetBrains AI subscription. JetBrains AI remains a suitable option for managing multiple models from one place, but it is no longer the only way to use agents within WebStorm.
A Protocol That Separates the Development Environment from the Agent
ACP operates as an open standard that separates the integrated development environment from the AI agent, similar to the role played by the Language Server Protocol, or LSP, in separating editors from language-support tools. Instead of changing editors to obtain a better tool for a particular language or task, developers can remain in WebStorm and choose the agent best suited to the task, whether that involves refactoring a React application or planning a software architecture.
WebStorm sends the agent the required context, such as files, diffs, and terminal output, while the agent returns actions that can be executed within the project, such as file modifications, tool calls, and shell commands. This allows developers to switch among a set of specialized agents while retaining WebStorm’s local indexing, navigation, and refactoring capabilities.
The importance of this approach, according to the article, stems from differences in agents’ strengths. In an independent test of design-to-code conversion tasks from Figma, agents achieved scores between 3.3 and 3.6 out of 5 for component structure, but their results in extracting design tokens fell to a range between 1.7 and 2.9. No single agent leads in every area, and the ranking of agents may change as new versions appear.
Integrations and Custom Agents
Instead of building a separate integration for every new agent, any agent that implements ACP becomes a first-class citizen within WebStorm. GitHub Copilot provides an ACP server, while Claude Code has an official ACP adapter, with other major agents joining the ecosystem.
- Registry and custom agents: Agents can be selected from the JetBrains registry, such as Junie, or a custom ACP-compatible agent can be connected through an acp.json configuration.
- Bring your own key and control the infrastructure: Because ACP is not tied to a specific provider, it allows agents to be directed to Azure, AWS, Anthropic, or local language models.
- Governance and data: Communication takes place directly between the development environment and the agent, helping teams keep data flows within the boundaries approved by their legal or regulatory authorities.
A Workflow from Design to the Browser
JetBrains provides an example of building a complex React dashboard using multiple agents. The work begins by selecting an agent specialized in front-end development. Figma Connect for WebStorm then brings the design context directly into WebStorm and sends the agent the relevant theme files and component specifications. The agent subsequently returns file modifications, and the ACP server applies them to the project.
The role is not limited to writing code. When asked to fix a console error and verify the solution in the browser, Chrome Connect, a built-in skill for the Chrome DevTools command line in WebStorm, gives the agent direct access to Chrome. It can read console logs and network requests, trace the error to its source, apply the fix, and confirm that the problem has disappeared. This integration requires a one-time setup and was developed in partnership with Google’s Chrome DevTools team.
The same idea can be applied to tasks such as creating Conventional Commits messages, preparing for code review, making changes to multiple files, and writing documentation. The difference is that the files, project, and current diffs remain in place when switching agents; only the agent changes.
What ACP Means for Teams
JetBrains believes ACP can align with AI governance structures at technology leaders. Teams can define approved providers, compliance requirements, and available agents, then make them available to developers without repeated individual setup. They can also build ACP-compatible agents preconfigured with knowledge of internal projects, coding standards, and private API documentation.
ACP is now available in WebStorm. It can be accessed from the Agents menu within JetBrains AI, where third-party agents appear alongside Junie and Quick Edit whenever they become available through the JetBrains AI platform. WebStorm’s ACP documentation explains the steps for connecting a custom agent or configuring BYOK.