Eclipse Theia 1.71 Release: News and Noteworthy

We are happy to announce the Eclipse Theia 1.71 release! This release ships 79 merged pull requests. In this article, we will highlight some selected improvements and provide an overview of the latest news around Theia.

For those new to Eclipse Theia, it is the next-generation platform for building IDEs and tools for the web or desktop, based on modern state-of-the-art web technologies. With Theia AI, part of the Theia platform, you can also build AI-powered tools and IDEs. For more details about Eclipse Theia, please refer to this article and visit the Theia website.

We are happy to share that the Eclipse Foundation Newsletter recently published an overview article on Eclipse Theia: The Eclipse Foundation tool platform in production, highlighting how Theia has progressed from “next-generation tool platform” to a foundation powering production tools and custom IDEs across industries, including recent releases such as Samsung’s Sokatoa and STMicroelectronics’ STM32CubeMX2. In a related adopter story, Samsung Sokatoa — GPU Profiling and Debugging for Android takes a closer look at how Samsung leveraged Theia’s modular and extensible platform to build a highly specialised, domain-specific tool that goes far beyond traditional code editing.

Eclipse Theia was also strongly represented at the recent Open Community Experience 2026 (OCX26) conference in Brussels with three dedicated talks, see the Theia at OCX 2026 section below for details and recordings.

The Theia project also releases a product, the Theia IDE. The Eclipse Theia IDE is a modern, AI-native, and open IDE for cloud and desktop environments, aimed at end users. The Theia IDE is based on the Theia platform and also includes advanced AI powered features. For more details, see the Theia IDE website.

If you are looking for a simple way to check out the new release, please download and install the Theia IDE, which is based on Theia 1.71.

Theia at OCX 2026

Eclipse Theia was strongly represented at Open Community Experience 2026 (OCX26), the Eclipse Foundation’s flagship event held in Brussels, with several dedicated talks covering the platform from different angles. If you want to catch up on the latest from the Theia ecosystem, the recordings are now available:

Eclipse Theia — News from the next-gen tools platform by Thomas Mäder (Castle Ridge Software) and Jonas Helming (EclipseSource) provides a broad update on the Theia platform, Theia AI, and the Theia IDE, covering recent developments in architecture, AI integration, and real-world adoption across industries.

AI in Action: The ultimate live demo with Theia AI by Jonas Helming (EclipseSource) is a slide-free, fully live demonstration of AI-native development with Theia AI and the Theia IDE. The session shows AI agents generating and modifying applications, automating testing and code review workflows, interacting with external systems, and even driving domain-specific tools such as AUTOSAR diagram editors, along with how capabilities, sub-agents, skills, and custom prompts can be used to shape agent behaviour.

Eclipse Theia in practice: Getting started and lessons from the field by Stefan Dirix (EclipseSource) takes a practical, experience-driven look at building custom tools on Theia. The talk walks through the minimal setup with the Theia CLI, the frontend/backend/RPC architecture, and shares earned lessons from production: e.g. using contribution points instead of patching core components, keeping a clean frontend-backend separation, packaging and deployment trade-offs, and when to choose Theia extensions over VS Code extensions.

Eclipse Theia 1.71: Selected features and improvements

In the following, we will highlight some selected improvements in the new release. As usual, we cannot mention all 79 improvements, however we will focus on the most notable changes as well as changes visible to end users. The corresponding pull requests are linked under the respective heading when applicable.

SCM History Graph

Theia 1.71 introduces a Source Control History Graph: a fully rendered commit DAG that lives inside the SCM view container. This new view enables extensions like the built-in VS Code Git extension to render a commit history with branch lane lines, ref badges, author information, tooltips, and context menus. The implementation includes a VS Code-compatible plugin API for SourceControlHistoryProvider, a paginated graph model, and a pure function for DAG lane computation supporting linear, merge, branch-out, octopus, and sibling topologies. Ref badges are deduplicated (local and remote refs collapse), and lane colors make it easy to follow branches visually. Simply open a workspace with a Git repository, and the graph is available within the Source Control view.

SCM History Graph showing branch lanes, ref badges, and commit entries with detailed tooltips

Workspace Trust for AI Features

All AI features in the Theia IDE are now gated by Workspace Trust. Until you grant trust to a folder, the IDE runs in Restricted Mode: chat input, inline completion, and AI commands are disabled, workspace-scoped prompt templates (.prompts/*.prompttemplate) are not loaded, workspace-scoped AI preferences (tool confirmation, model aliases, per-agent settings) are ignored, and external images embedded in AI chat markdown are blocked. The AI views and menus stay visible so users can discover the features, but a clear “Manage Workspace Trust” prompt is shown until trust is granted. A follow-up fix suppresses a spurious restart dialog that was previously triggered by trust-related preference changes.

This complements the broader Workspace Trust framework that protects tasks, debug sessions, extensions, and MCP servers. The new dedicated Workspace Trust documentation page covers the details, including the trust dialog, status bar indicator, and security.workspace.trust.* settings.

Theia AI Improvements

Theia AI is the framework within the Theia platform for building AI-powered tools and IDEs. The Theia IDE and its AI Coding features, including Theia Coder, are built on top of Theia AI, but adopters can also use it to build their own domain-specific AI solutions. Most of the improvements listed below are directly visible to Theia IDE end users, for example the new reasoning level selector, the PR Review Agent, token usage indicators, and updated default models. At the same time, they are also valuable for Theia AI adopters, who benefit from the same provider-agnostic abstractions, customizable tools, and white-labeling support in their own products.

Provider-Agnostic Reasoning Level Selector: The previous thinking mode settings (a binary enabled flag plus a budget-tokens field modeled after Anthropic’s original extended-thinking API) have been replaced with a provider-agnostic reasoning level selector. The new ReasoningSettings abstraction supports levels from off through minimal, low, medium, high, to auto. Each LLM provider declares a ReasoningSupport capability on its model description, and a compact UI selector appears next to the chat mode selector, visible only for reasoning-capable models. Both effort-based APIs (Anthropic adaptive thinking, OpenAI, Gemini 3) and budget-based APIs (Anthropic legacy, Gemini 2.5) are supported. Precedence at request time is: session override, then the new ai-features.reasoning.defaults preference, then the model’s declared default.

Provider-agnostic reasoning level selector in the Theia AI chat view, allowing users to choose the reasoning effort for reasoning-capable models

PR Review Agent with UserInteractionTool: A new @pr-reviewer agent conducts structured pull request reviews by orchestrating the GitHub agent, workspace tools, and agent delegation. As part of this, a generic UserInteractionTool is introduced that allows any agent to ask the user structured questions with labelled option buttons, an optional free-text fallback, and navigable links to open files and diffs in the workspace. The following video shows the new agent in action, for more details refer to this article and watch the demo below.

Token Usage Indicator and Warnings: A per-session token usage indicator has been added to the chat view, showing current conversation context size via a compact circular progress ring next to the send button. The ring fills proportionally with usage, turning yellow at 90% and red at 100%+, and the chat input border tints accordingly. A hover tooltip provides a detailed breakdown of input, output, cached, and total tokens. Additionally, a new token usage warning system can alert users with a dismissable notification when a session’s total token usage crosses a configurable threshold, offering quick actions to compact the session or start a new chat. Both features are controlled via preferences under ai-features.chat.tokenUsageIndicator and ai-features.chat.tokenUsageWarning.

Token usage indicator in the Theia AI chat view, revealing detailed information such as the conversation context tokens on hover

When the configured warning threshold is reached, a user notification informs the user to summarize the session, start a new chat, or update the settings

Updated Default AI Models: Default models have been bumped across AI packages — Claude Opus 4.7 is now the preferred Anthropic model, OpenAI defaults have been updated to the GPT-5.5 family including GPT-5.4-nano, and older models (GPT-4.1, GPT-4o) have been removed.

Product Name Variable for White-Labeling: A new #productName AI variable resolves to the configured application name, and all built-in prompts and agent descriptions now use this variable instead of hardcoded “Theia” or “Theia IDE” references. Adopters building custom tools on Theia AI can set their application name and have all AI-generated responses refer to their product correctly.

Customizable Copilot OAuth Configuration: Downstream adopters can now override the GitHub OAuth app configuration and authentication dialog messages for the Copilot integration via DI rebinding, enabling products to use their own GitHub OAuth app and product name.

Shell Command Descriptions in Tool Confirmation: The shell execution tool now accepts an optional description parameter from the LLM, which is displayed in the approval confirmation dialog. This lets users make more informed approve/deny decisions when an AI agent requests to run shell commands.

Dev Container Improvements

Dev Container support receives several improvements in this release. The highlights include SSH credential injection and Git identity sharing, where the container automatically receives an isolated SSH directory with key forwarding, host .gitconfig is injected with SSH signing key rewriting, and a login shell is configured with SSH agent. Auto-reopen in last active dev container reconnects to the previously used container on application restart. A new “Reopen in Container” suggestion notification appears when opening a workspace that contains a devcontainer.json. A Rebuild Container command removes the old container and creates a fresh one while preserving the devcontainer file context. And a new “Attach to Running Container” command lets you connect to any already-running Docker container without needing a devcontainer.json. Beyond these user-facing features, the release also fixes variable resolution for embedded strings in devcontainer.json, ensures all mount options are passed through correctly, and fixes JSON quoting when injecting preferences.

ANSI Color Support in the Output View

The Output view now parses and renders ANSI SGR escape sequences, replacing the previous raw escape code display with properly styled inline decorations. Colors, bold, italic, and underline styles render correctly via Monaco editor decorations, using CSS classes that reference existing terminal ANSI color variables. Build tools, test runners, and language servers that emit ANSI-colored output now look as intended across both light and dark themes.

ANSI color and style rendering in the Theia Output view, with colors, bold, italic, and underline styles displayed inline

Test Extension Support Improvements

Theia 1.71 fixes several long-standing issues that prevented VS Code test extensions like Python unittest and Playwright Test from working reliably, and improves the testing UX:

• Workspace folder reference equality is now preserved in workspace.getWorkspaceFolder(), fixing extensions that use the folder object as a Map key.
• Root-level test discovery is now correctly triggered by calling resolveHandler(undefined) when a controller registers, matching VS Code behavior.
• Test state updates are no longer dropped when a run ends — TestRun.end() now flushes pending change batches.
• TestItemCollection.replace() no longer deletes newly added items that share IDs with old ones.

On the UX side, the Test Output panel now opens automatically when a test run starts, a new “Show Test Output” toolbar button has been added to the Test Runs view, and double-clicking a test run or item also opens the output panel — falling back gracefully to failure messages or a placeholder when no output is captured.

Git & SCM Improvements

Beyond the new History Graph, the SCM experience has been further refined.

Repository Context Menus and Worktree Grouping: Right-click menus in the Repositories view now correctly route to the scm/sourceControl menu path, the ... button is properly wired to the scm/repository menu, and worktrees are correctly nested under the root of their parent repository. This also enables the Worktrees submenu contributed by the Git extension.

External Worktree Removal Detection: When a worktree is deleted outside of Theia (via CLI, file explorer, or parent repo closure), the Repositories view now correctly detects the removal and updates its state. This closes three gaps in the SCM wiring around parent-child cascade, plugin-side registry cleanup, and stale selection handling.

Core Platform Improvements

Improved Start-Up Performance Logging: Start-up performance logging has been expanded to give a clearer picture of application initialization time, with measurements for individual configure() calls, Electron main process timing, pre-container bootstrap phases, and updated performance scripts. FrontendApplicationContribution.initialize() now supports async initialization.

RPC Promise Rejection Guarantees: Several gaps in the RPC framework have been fixed to ensure client promises are properly rejected when websocket connections drop or the write buffer overflows, preventing promises from hanging indefinitely. A new errorHandling option on Emitter supports propagation, logging, or custom callback modes.

Connection Container Cleanup: All services in backend connection containers are now unbound when the socket definitively closes, allowing @preDestroy hooks to run and clean up allocated resources.

Standard Fetch API: The Node.js request service has been migrated from the legacy http(s)-proxy-agent to the standard fetch API with undici for proxy support, reducing dependency surface area.

Autosave on Window Focus Loss: When the autosave mode is set to onFocusChange, files are now saved both when the tab loses focus and when the application window loses focus, matching VS Code’s behavior.

VS Code API Bump to 1.116.0: API compatibility has been bumped to VS Code 1.116.0 with updated NLS metadata.

Electron 39.8.7: Electron has been upgraded from 39.7.0 to 39.8.7, fixing an intermittent crash on Windows when interacting with the splash screen during startup. The Electron window close behavior on Windows has also been fixed — closing via the title bar “x” button no longer shows a stale “Offline” state.

Additional Improvements

The 1.71 release includes many more noteworthy changes beyond the scope of this article, e.g.:

Chat view accessibility: the AI Chat view has received keyboard accessibility improvements. Toolbar items are now in the tab order, welcome page cards show action buttons on keyboard focus, and the problematic Shift+Tab keybinding for “Cycle Chat Mode” has been replaced with Ctrl+M to avoid trapping keyboard users.

Copy in chat view (Electron): context menu copy in the AI Chat view now correctly copies the whole message when there is no DOM selection, fixing a regression where Electron’s native role bypassed the custom copy handler.

Terminal copy via context menu: right-click → Copy in the terminal now works correctly in both browser and Electron by using xterm’s virtual selection instead of relying on DOM text selection.

Navigator copy/paste safety: the file navigator copy/paste handler has been hardened to prevent invalid clipboard content (bare filenames instead of URIs) from triggering unintended filesystem traversals.

File watcher improvements: plugins can now watch files outside the workspace using createFileSystemWatcher, and an optimization avoids creating overlapping watchers when the workspace root watcher already covers child paths.

Localization fix: a preload import issue that caused some strings not to be properly translated after startup has been resolved.

Agent icons: all built-in AI agents now have proper distinct icons, and the AI Configuration view correctly displays them.

For a complete overview of all changes in this release, please refer to the 1.71 milestone. All these features and improvements (in total 79) were the result of one month of development. Eclipse Theia follows a monthly release schedule. We are looking forward to the next release due next month, stay tuned! To be notified about future releases, follow us on LinkedIn or follow Theia on Twitter and subscribe to our mailing list.

If you are interested in building custom tools or IDEs based on Eclipse Theia, EclipseSource provides consulting and implementation services for Eclipse Theia, for AI-powered tools, as well as for web-based tools in general.

Furthermore, if you want to extend Theia, Theia AI or the Theia IDE with new features or strategically invest into the project, EclipseSource provides sponsored development for Theia, too. Finally, we provide consulting and support for hosting web-based tools in the cloud.

👉 Get in contact with us, to discuss your use case!

👉 Subscribe to our YouTube channel