Eclipse Theia

AI coding agents are helping developers produce more code than ever, but that also makes code review a growing bottleneck.

We therefore introduce the PR Reviewer agent in the Theia IDE: an AI-assisted code review workflow that supports human reviewers instead of replacing them. The agent helps with tedious review tasks such as preparing the workspace, navigating changes, drafting findings, and posting comments, while the reviewer stays in control of the final outcome.

Eclipse Theia 1.70 is out, just in time for OCX next week: AI Coding in the Theia IDE has graduated from beta, alongside major improvements to agent workflows, Git/SCM, terminal history, search, debugging, and platform compatibility.

The Eclipse Theia Community Release 2026-02 is here, bringing together 167 improvements from Theia 1.67 and 1.68 into a stable milestone for adopters. From GitHub Copilot integration and Agent Skills to shell execution, stronger agent workflows, AppTester automation, and the new Terminal Manager, there is a lot for Theia builders and users to explore in this release.

Our new demo video shows Theia Coder across a complete workflow, from app generation and code updates to planning and automated E2E testing. It also introduces the Capability concept for enabling workflows like testing, shell access, and GitHub-related tasks more easily.

Our new article revisits domain-specific AI extensions in VS Code and examines how the editor’s native AI APIs have matured across language models, tools, chat participants, renderers, and MCP integration. It is a useful read for teams deciding between extending VS Code and building fully custom AI-enabled tools on platforms like Eclipse Theia AI.

Eclipse Theia 1.69 introduces Agent Capabilities - a new abstraction for toggling agent skills without the complexity - alongside smart shell command permissions, Claude Code integration improvements, Thinking Mode configuration, MCP Roots support, and extensive chat UX polish. 88 merged PRs in total.

Eclipse Theia 1.68 lands with Copilot integration, a significantly improved Architect Plan Mode, Skill Support, Shell Execution,an upgraded Coder Agent Mode (Next), improved UI Testing, and more. Read the highlights (and note it’s the RC for the 2026-02 community release).

The 5 most common bad habits in AI coding. Spoiler: it's not about your prompts. It's about what happens after that first generation.

We found one pattern connecting all five. Once you see it, you can't unsee it.

AI coding tools are transforming software development — but the conversation focuses on productivity while ignoring the human cost. This video explores why developers increasingly report feeling drained, unproductive, and even useless despite higher output.

MCP makes it easy to plug AI agents into everything, but that convenience can backfire when tool lists and descriptions flood the context window. This post explains why “more tools” can mean worse outcomes and shares concrete ways to keep agents lean and reliable.