Using several AI coding tools creates a coordination problem: someone still needs to define the work, manage dependencies, keep changes separate, review the results, and decide what is ready to merge.
Coterie is a Windows desktop app that brings those steps into one workspace. It coordinates supported official AI coding CLIs through distinct roles: an Architect plans the work, Builders implement tasks, and independent Reviewers inspect the proposed changes.
Start with a structured plan
A development goal often contains several tasks with different dependencies. A frontend change might depend on an API update, while tests or an unrelated component can be developed separately.
Coterie’s Architect role turns the goal into structured tasks and identifies how they relate. This gives the implementation agents a defined scope and makes the execution order easier to follow.
The plan provides the foundation for the rest of the workflow: what needs to change, which tasks can proceed independently, and what evidence should be available when the work is reviewed.
Run independent implementation tasks in parallel
Developer agents work in separate Git worktrees. Each worktree gives a task its own working directory and branch, helping keep proposed changes separate from the main checkout.
Independent tasks can run simultaneously within the project’s configured capacity. Tasks with prerequisites follow their dependencies.
This structure helps developers manage parallel work while retaining a clear view of which agent owns each task and which changes belong together.
Review changes with a different model
Coterie separates implementation from review. A different model reviews the proposed work against the task requirements and available evidence.
The purpose is to introduce an independent assessment before the workflow advances. Reviewers can inspect the changes, identify missing requirements, and request corrections.
Model diversity adds another perspective, while project checks provide concrete evidence about the implementation. Both contribute to the decision about whether a change is ready.
Understand what “verified” means
In Coterie, verification is tied to the task requirements, reviewer findings, and the checks configured for the project. Depending on the repository, those checks can include tests, type checking, linting, or builds.
Completion reports bring the results together so developers can inspect what changed, which checks ran, and what still needs attention.
Passing checks establishes that the change met those particular checks. Developers should assess whether the coverage and acceptance criteria are appropriate for the work being delivered.
Follow the work in one workspace
Coterie includes a visual hierarchy map, dependency views, a task board, change inspection, and completion reports.
These views help answer practical questions throughout development:
- Which tasks are running?
- What is waiting on a dependency?
- Which changes are under review?
- What needs correction or approval?
- What evidence supports the completed work?
Project automation settings govern how tasks advance and where human approval is required.
Who is Coterie for?
Coterie is designed for software developers, technical leads, and independent builders who use AI coding CLIs and want a more structured way to coordinate their work.
It works with supported official coding tools and their existing provider access. Provider account requirements, usage limits, and supported CLI requirements continue to apply.
Request private beta access
Coterie is available as a free private beta for Windows 10 and Windows 11. Access is provided on request.