Peter Steinberger's September 22nd release added Charm Hyper, GitKraken AI and Bifrost, alongside a JavaScript plugin system for new integrations.
By [RuntimeWire Staff](https://runtimewire.com/author/runtimewire-staff)
· Published
Primary source: [X](https://x.com/steipete/status/2102672670689275965)
Why it matters #
CodexBar is turning a focused usage meter into an extensible integration layer. Its 80-provider tally shows breadth, while the JavaScript plugin system could let contributors help maintain that breadth as AI coding services multiply.
Peter Steinberger (@steipete) expanded CodexBar's reach across the crowded market for AI coding tools: in a September 23rd post on X, he listed support for TypeSafe, Nous Portal, Muse Code, CodeRabbit, Replicate, Hugging Face, Pi, Helmcode, v0, Charm Hyper, GitKraken AI and Bifrost. The project's 0.65.0 changelog, dated September 22nd, specifically records three additions - Charm Hyper, GitKraken AI and Bifrost - and puts the provider catalog at 80.
The gap between the two lists is worth keeping straight: Steinberger's post names a dozen supported services, while the release notes identify the three additions in that version. They do not say that all 12 arrived together. The practical news is a broader tracker, with a new way for the project to absorb providers without treating each one as a bespoke addition to the app's core.
A founder's utility becomes a platform for integrations
Steinberger built PSPDFKit, a developer tools business founded in 2011 that later became part of the document software company now known as Nutrient. Nutrient's account of PSPDFKit's history describes its start as an effort to bring a magazine to the iPhone with a PDF viewer. CodexBar is a different kind of product: a small utility for making the limits and usage of AI services visible while people move among them.
The CodexBar repository describes a macOS menu-bar app and Linux desktop and command-line tools. Depending on the provider, it displays usage windows, reset times, credits or spending. That places it beside the coding tools themselves, not in competition to generate code. Its value is coordination: users can see several services' allowances in one place rather than checking each service separately.
That job grows as developers add subscriptions and accounts to their working set. A coding assistant can be available while its quota is not; a credit balance or reset window can determine whether a long task stays on one service or moves elsewhere. CodexBar makes those constraints visible. Its 80-provider count describes integration breadth, not the number of active users, paying customers or services used by a typical person.
Plugins move the maintenance burden outward
Steinberger also said the app now has a plugin system that lets developers write providers in JavaScript. The change matters because every provider has its own way of exposing usage. CodexBar's documentation describes integrations that can draw on existing OAuth sessions, API keys, browser cookies, local files or command-line tools. The details differ by service, and those connections can break when a provider changes its interface or authentication flow.
Moving new integrations into JavaScript plugins gives contributors another route to add coverage without adding every provider implementation to the native application. It does not make those connections maintenance-free: a plugin still needs a reliable data source, and access through cookies or local credentials carries practical security considerations. CodexBar says it reuses existing provider sessions and does not store passwords; users still need to decide which sources to enable.
The plugin design also changes the project's growth equation. A tracker that supports more services becomes more useful to people whose work spans several tools, while each newly supported provider adds a continuing compatibility obligation. Letting contributors write plugins could widen the pool of people who can extend CodexBar. Whether the approach keeps integrations working as services change will depend on the quality and upkeep of those plugins, not the headline provider count alone.
CodexBar remains an open-source project on GitHub, where the repository showed about 21,800 stars and 2,000 forks in the supplied snapshot. Those figures show substantial public interest, but they do not establish usage, revenue or a commercial operation. The project has no funding figure in the available reporting; Steinberger's earlier work at PSPDFKit should not be mistaken for financing of CodexBar.
For Steinberger, the bet is a familiar one in a new setting: build a focused tool around friction developers encounter in day-to-day work, then let the product's usefulness grow with the surrounding software. CodexBar's documented job is to make usage limits visible. Its provider catalog and plugin system extend that utility across the AI coding tools people choose to use.