JetBrains' ThinkRail Draws 1,075 Stars in a Day, Supports Only One Agent and Sends Usage Reports by Default
Software / news
JetBrains' ThinkRail Draws 1,075 Stars in a Day, Supports Only One Agent and Sends Usage Reports by Default
The Apache-2.0 IDE built on the pi coding agent is a JetBrains incubator project at version 0.2.0; its README and a third-party review disagree on Intel Mac support.
JetBrains' ThinkRail, an Apache-2.0 desktop IDE built around the open-source pi coding agent, gained 1,075 GitHub stars in one day, according to GitHub's trending list on October 28, and stood at 4.8k stars with 186 open issues when its repository was read for this story. The README labels it a "JetBrains incubator project", and its own install examples reference versions 0.1.2 and 0.2.0-nightly.10.
Its README says basic usage and error reports are always sent, with extra usage sharing opt-in and --no-analytics available to switch it off. That line sits under Analytics, not at the top.
What ThinkRail is, according to its README
The JetBrains/thinkrail repository describes the project as "the agentic IDE that gets better every time you use it." It wraps pi and gives each task its own git worktree and branch, with pull request status and merge actions in the app.
The workbench has Monaco editor tabs, a file tree with git change marks, terminals and anchored review comments. A workspace can hold several concurrent pi sessions with live token and cost readouts and subagent delegation. A graph of SPEC.md files is read and maintained by the agent. MCP servers prompt for approval before any call that is not read-only.
Sign-in works with JetBrains AI or any provider configured through pi. The UI ships in English, Japanese, Simplified Chinese and Spanish. The architecture is a Bun.serve engine host, a typed wire contract and a React 19 client.
Who built it and when
A review on aicoolies.com credits JetBrains InnovationHub and dates the launch post, "Introducing ThinkRail", to August 19, 2026. It cites version 0.2.0 in a blog post dated October 18, 2026. Those dates come from that third-party page, and JetBrains' own posts were not checked for this story.
The same page lists the app as free, with no accounts or paid tiers of its own. Model costs depend on the provider, whether a JetBrains AI plan or a user's own key. It also flags "Concerns reported" about telemetry without saying what they are.
Pi only, and a crowded JetBrains shelf
The sharpest criticism so far comes from Java Architecture Diary, a blog, in a post dated September 10. It notes that Orca runs Claude Code, Codex, OpenCode and Pi side by side, while ThinkRail supports only Pi. The author wrote that "ThinkRail must justify why it deserves adoption despite its single-agent constraint."
The post traces JetBrains' agent products: AI Assistant in late 2023, Junie in early 2025, a Junie and AI Chat merger announced in December 2025, Fleet ending updates in late 2025 and returning as Air, and Air entering public preview in March 2026. It argues that JetBrains now maintains Air, which already supports multiple agents, while incubating a pi-only tool. Its verdict: "The pattern: ship fragmented entry points, then slowly consolidate."
That is one blogger's reading, not a JetBrains position. The README says only that "full extension support" is next.
Platform requirements and a conflict between sources
| Item | README | aicoolies.com |
|---|---|---|
| macOS | 14+, Apple Silicon or Intel | Apple Silicon only |
| Windows | x64 | x64 |
| Linux | x64 and ARM64, glibc 2.38+ | x64 and ARM64, Ubuntu 24.04+ |
The sources disagree on Intel Macs, and the README is the primary document. The desktop build needs Bun 1.4.4, Node.js 22.19 or later, git on the PATH and an authenticated pi provider. The macOS CLI is signed but not notarized, and Linux artifacts are unsigned, so Gatekeeper may block some older or locally built macOS disk images.
What the star count does and does not show
A day of 1,075 stars says people bookmarked a JetBrains repository. It does not say how many run it, and the 186 open issues against 604 commits on main suggest a project still sorting out its basics. Agent tooling that sends reports by default has been a pattern: Hallmark and Impeccable drew 128,000 stars while only one README documented its telemetry, and ThinkRail at least documents its own.
Teams that hold code under strict data rules should read the analytics section before installing. Hosted agent backends have also been breached: OpenCode and Warp both lost server-side data within 10 days. The next marker is the extension support the README promises, which has no date.
Sources
More in Software
- 01Python 3.15 Ships a Built-In sentinel After Five Years of PEP 661The new type replaces the object() idiom with a one-line constructor that survives copying, pickling and type checking, but a Python 3.15 floor is the price of using it.
- 02Lightpanda 1.0 Passes 80% of Chrome's Web Tests, by Its Own CountThe Zig-built headless browser left beta on Oct. 2, and its own benchmarks put it ahead of two Rust rivals on a 940-page crawl, with no hardware listed.
- 03Claude-Mem Hits 107,900 Stars, and Its Own Pages Disagree on LicenceThe Claude Code memory plugin has gained about 35,500 stars since May, but a May write-up says AGPL-3.0, the repository now says Apache-2.0, and the README describes an optional cloud sync that earlier coverage did not.
- 04Jepsen Found 96 Issues in ParticleDB 1.1.4 and Tells Users to Halt DeploymentKyle Kingsbury’s 31 October report, funded by vendor Orkes, finds lost writes and ignored constraints in healthy single-node runs.