Wazear

Wazear runs locally on desktop computers and coordinates a sequence of AI agents that each perform a specific step in a development pipeline. One agent generates a deliverable—such as a plan, architecture, or code—while a paired reviewer evaluates that output against the original specification, and a final QA stage examines the complete result before handing it to a fixer for remediation. The system is built on Node.js and can use the free Qwen Code tier or any compatible API key (e.g., OpenAI, Gemini, OpenRouter).

The default configuration includes eight agents—Planner, Planner Reviewer, Architect, Architect Reviewer, Programmer, Programmer Reviewer, Fixer, and QA—arranged so that every producing step has an opponent reviewer, providing two quality gates: internal review and user‑acceptance testing. Users can edit, remove, or replace any agent, define custom pipelines without programming, and add or modify command‑blocking guardrails to enforce safety policies.

Wazear is positioned as a structured alternative to ad‑hoc chatbot prompting, emphasizing clear role boundaries, selective information exposure for each agent, and the ability to tailor the workflow to specific tasks. It is currently available for Linux and Windows, with macOS support planned, and is offered under an experimental maturity level.