Skip to content
ReviewRamble

The weekly brief

AI, reviewed, in your inbox.

  • New reviews of tools, models, MCP servers and repos, with the verdict up front
  • Re-checked prices and ratings when something changes
  • New analysis pieces as they are published

Models, agents, MCP servers and the tools actually worth your time.

One email a week. Unsubscribe with the link in any brief. How we use your email

OpenIDEA

An IntelliJ-style IDE rebuilt on a Rust core, with Claude Code, Codex and 13 other agents in worktree tasks and an MCP server.

Reviewed by ReviewRamble editors · Updated

View on GitHub Open source

Share this page:XLinkedInFacebookRedditWhatsAppEmail

Our rating

3.8/5

OpenIDEA is the first serious attempt at an IntelliJ-shaped home for terminal coding agents, and at a month old it is too young to trust with a main codebase.

What is OpenIDEA?

OpenIDEA copies the way IntelliJ IDEA works, its keymap, tool windows and search-everywhere, into a native desktop app built with Tauri on a Rust core. It ships an editor, code insight through language servers, a debugger, Git, a terminal, and database and HTTP clients for more than 60 languages, with builds for macOS, Windows and Linux on x64 and arm64.

The AI side has two halves. An Agents tool window starts command-line agents you have already installed and signed in to (Claude Code, Codex, Gemini CLI, GitHub Copilot, Cursor, OpenCode, Qwen Code, Kimi, Goose and others, 15 built in), each conversation as a task in its own Git worktree or in your checkout, with diffs to review. A built-in MCP server, off by default, lets agents running outside the IDE use its search, symbol lookup, Problems view, run configurations and Git status.

It is published by l0ng-ai under the Apache 2.0 licence and is not affiliated with JetBrains. The repository was created on 21 October 2026; v0.1.49 shipped on 19 November 2026.

Is OpenIDEA worth it?

OpenIDEA matters less as an IntelliJ clone than as a place to run coding agents with an IDE's tools around them. It brings no model of its own: it launches the agent CLI you already pay for over the Agent Client Protocol (Claude Code and Codex through their own protocols), gives each task a Git worktree on an agent/<task> branch, shows the edits as diffs per task or per turn, and lists tasks by what they need from you. With its MCP server switched on, every task also gets the IDE's index, language-server lookups and run configurations as tools, which is more than a bare terminal agent sees.

The speed claim is the headline, and the README is careful with it. Opening freqtrade, a Python project of about 770 files, OpenIDEA reached a settled state in 1.1 seconds against 13.2 for IntelliJ IDEA 2026.2.3 and used 169 MB of memory against 2,011 MB, on a 32-core Windows 11 machine. The same table says IDEA analyses Python more deeply, so the two are not doing the same work; the gap reflects a language-server design that cannot match JetBrains' own semantic model, which the docs admit for features such as structural search.

Two defaults deserve attention. The MCP server binds to 127.0.0.1 with a bearer token and refuses other origins, but it can be pointed at 0.0.0.0, and the token then guards your projects on that network. More surprising, once an agent has run a task, new tasks start in its approve-everything mode (Claude Code's bypass permissions, Codex's Full Access) unless you change Settings to ask for approval. Agents and language servers are blocked in projects you have not marked as trusted.

The cost is maturity. Version 0.1.49 is the 46th release on GitHub since 22 October, the history shows more than 3,000 commits since 20 October, and 208 issues and pull requests are open. The plugin API, sandboxed WebAssembly components, is labelled a preview, and the Windows packages are not code-signed, so SmartScreen warns on first run. It suits someone willing to report bugs; for a team that lives in IntelliJ, it is a side tool for agent work at best for now.

Where it earns its place

  • A backend developer on a laptop who finds IntelliJ IDEA slow on a mid-sized Python repository

    Keep IntelliJ shortcuts and tool windows but get a lighter editor

    Install the release for the platform, import JetBrains settings, and open the project; the README's own test shows a settled window in about 1.1 seconds. Free; expect fewer deep Python inspections than IDEA.

  • A solo developer paying for Claude Code and Codex

    Run two agents on separate fixes at once and review each before merging

    Start each as a New Agent Task in its own worktree, answer permission prompts as they come, then read the per-task diff and merge the agent/ branch with the Git tools. Agent usage is billed by Anthropic and OpenAI as usual.

  • A developer who runs Claude Code in a terminal next to the IDE

    Let the terminal agent use the IDE's symbol search and Problems view

    Enable the MCP server in Settings, copy the stdio config, and add it with claude mcp add; the bridge holds no token and survives IDE restarts, as long as OpenIDEA is running.

Before you commit

Setup
Download a .dmg, Windows setup, MSI, portable zip, .deb, .rpm or AppImage from GitHub Releases (a China mirror is on CNB). macOS 12 or later; Windows needs WebView2. Install each agent CLI and sign in once; OpenIDEA finds it on PATH.
Integrations
Agent Client Protocol agents plus a registry of more; Claude Code stream-json and codex app-server; an MCP server over stdio or HTTP on port 26371; LSP and DAP for languages; sandboxed WebAssembly plugins, including an official Ollama or OpenAI completion plugin.
Data and privacy
Runs locally; agents use your own credentials and send code to their providers. Agent environment variables and MCP secrets go to the OS credential store. The MCP server is off by default and loopback-only with a bearer token when on. Telemetry practices are not published in the docs we read.
Cost at scale
The IDE is free; the bill is the agents. Several parallel worktree tasks multiply Claude Code, Codex or API usage, and OpenIDEA shows context use, usage limits and session cost when the agent reports them.
Limits
Worktree tasks need a Git repository with at least one commit; agents are off in untrusted projects; idle resumable agents stop after five minutes; features that need a full semantic model, such as structural search, are missing for some languages.

What’s good

  • Runs 15 agent CLIs as worktree tasks with per-turn diffs
  • Built-in MCP server gives agents the IDE's index and run tools
  • Measured 169 MB against 2,011 MB for IDEA on the same project
  • IntelliJ keymap and JetBrains settings import ease the switch

What’s not

  • 0.1.x after one month, with 208 open issues and pull requests
  • New tasks default to approve-everything after the first run
  • Language-server code insight is shallower than JetBrains'
  • Windows builds unsigned; plugin API still a preview

Best for

Developers who like IntelliJ's way of working and want to run Claude Code, Codex or other agent CLIs in parallel worktrees with diffs and IDE tools, and can live with an early release.

OpenIDEA alternatives

OpenIDEA: common questions

Is OpenIDEA worth it?
OpenIDEA is the first serious attempt at an IntelliJ-shaped home for terminal coding agents, and at a month old it is too young to trust with a main codebase. We rate it 3.8 out of 5.
How much does OpenIDEA cost?
Free under the Apache 2.0 licence; agents run on your own subscriptions or API keys, which OpenIDEA does not proxy or meter.
What are the downsides of OpenIDEA?
0.1.x after one month, with 208 open issues and pull requests; new tasks default to approve-everything after the first run; language-server code insight is shallower than JetBrains'; windows builds unsigned; plugin API still a preview.
Who is OpenIDEA best for?
Developers who like IntelliJ's way of working and want to run Claude Code, Codex or other agent CLIs in parallel worktrees with diffs and IDE tools, and can live with an early release.
How much does OpenIDEA cost for a team or in production?
The IDE is free; the bill is the agents. Several parallel worktree tasks multiply Claude Code, Codex or API usage, and OpenIDEA shows context use, usage limits and session cost when the agent reports them.
Is OpenIDEA safe for company data?
Runs locally; agents use your own credentials and send code to their providers. Agent environment variables and MCP secrets go to the OS credential store. The MCP server is off by default and loopback-only with a bearer token when on. Telemetry practices are not published in the docs we read.
What are the best OpenIDEA alternatives?
The closest alternatives we have reviewed are Daintree, T3 Code and OpenCode.

Answers checked Nov 20, 2026.