Zide opens beta for a native AI workspace spanning code, Git, and CI

Why it matters

Zide is betting that the durable business in AI coding sits above individual models, in the workspace that controls context, review, Git, and delivery.

Zide opened its public beta on August 24, giving founder Jim Bourke's native desktop workspace its first broad test against an increasingly crowded field of AI coding tools.

The release puts code editing, Git, issues, pull requests, continuous integration, terminals, worktrees, and AI agents inside one application for macOS, Windows, and Linux. Zide says its agent can inspect repository files, run approved commands, make changes across multiple files, and return those edits as reviewable diffs.

Bourke comes to developer tooling through a less predictable route. He founded the radio-controlled aircraft community RCGroups in 1996 and spent years building flight-simulation software at Knife Edge Software. He joined Knife Edge as a contractor around 2001, became general manager, bought the business in 2012, and later sold its RealFlight product to Horizon Hobby in 2021, according to a history published by Model Aviation. Zide's terms identify Knife Edge Software LLC as the legal entity doing business as Zide Software.

That history helps explain Bourke's product thesis. Flight simulators have to keep a large set of interacting systems visible and controllable. Zide applies a similar approach to software development, where the agent, source files, repository state, issue tracker, test output, and deployment checks can all affect whether a change is safe to ship.

"Developers shouldn't have to re-explain themselves to a new AI agent every time they switch apps," Bourke said in the announcement.

The workflow is the product

Most AI coding companies began with a narrow surface: an editor extension, an AI-first editor, a command-line agent, or a cloud environment. Those boundaries have been dissolving as agents take on longer tasks and developers spend more time supervising changes across repositories, terminals, tests, and pull requests.

Zide is placing that supervisory work at the center of the application. A developer can start from an issue, add files or diffs to an agent's context, review the proposed change, stage it, open a pull request, and monitor CI without moving the task through several separate interfaces. GitHub, GitLab, and Bitbucket integrations bring repository-hosting functions into the workspace, although issue support currently covers GitHub and GitLab.

The feature set includes Plan mode, which lets an agent investigate before editing; QuickTree, which creates a Git worktree for a new agent session; and Crosscheck, which sends one model's work to another model for review. Developers can also set tool permissions and require approval before sensitive actions run.

Those controls matter because a wider agent context also gives the software a wider field in which to make mistakes. Zide's answer is to keep changes, repository state, terminal activity, and CI results visible to the developer. The design treats human review as a continuing part of the loop instead of a final inspection after an agent finishes.

Zide says its desktop application is built with Tauri and Rust rather than packaged as a browser-based wrapper. That architecture supports Bourke's pitch for a persistent workspace across large projects and multiple repositories, though the public beta will have to establish how the application performs under that load.

Model choice becomes a distribution strategy

Zide is also declining to make a single model provider the center of the product. Zide Assist supports server-based models, users' own API keys or subscriptions, compatible local models, and OpenRouter connections. Zide's website names Anthropic's Claude, OpenAI's Codex, Google Gemini, and DeepSeek among its options.

That flexibility gives Zide a hedge against a market where model quality, pricing, and availability can shift faster than desktop software buying cycles. It also positions the application as the persistent layer above the models: developers can change the engine while keeping their repositories, sessions, layouts, and review process in place.

Zide is hardly alone in making that argument. Cline promotes broad provider choice, local models, terminal access, and multi-file changes. Cursor has pushed its editor toward parallel agents and Git worktrees; Cursor 2.0 introduced an interface designed around running multiple agents at once. Cursor's scale also shows the size of the opportunity Bourke is entering: Cursor said in November 2025 that it raised a $2.3 billion Series D at a $29.3 billion post-money valuation.

Zide's wager is narrower and practical. Developers may want an integrated agent environment without handing the entire workflow to one repository host, model vendor, or cloud development platform. Supporting GitHub, GitLab, and Bitbucket gives Bourke a route into teams whose infrastructure was decided long before coding agents arrived.

A low-cost beta with a long software license

Zide is free to download for public repositories. According to the launch announcement, the $10-per-month Solo plan adds unlimited private repositories, connections to the three supported repository hosts, and monthly Zide Assist credits. The $20-per-month Standard plan adds the full model catalog, bring-your-own-key access, and compatible local models.

The pricing structure also borrows from traditional desktop software. Customers who maintain a paid subscription for 12 consecutive months can qualify for a perpetual license to the eligible version installed at that point. Continued updates, AI services, and credits still require an active subscription.

That arrangement fits a founder whose career predates software's shift toward subscriptions and usage-based AI billing. It gives developers an ownership proposition while preserving recurring revenue for the services that carry ongoing model costs.

Bourke has built communities, simulators, and software businesses around technically demanding users before. Zide now has to prove that those users want a new desktop command center, rather than another agent inside the tools they already use. The beta's strongest premise is straightforward: as agents write more code, developers need a clearer place to see everything around that code before it ships.