imlazy is a "token-efficient, tier-routed" agent orchestrator for AI coding assistants (Claude Code, OpenCode, Codex CLI). Its core goal is to load only the necessary workflow context based on the actual complexity of the task, thereby saving tokens on everyday simple tasks while maintaining full engineering rigor on complex tasks.
Purpose and Positioning It automatically classifies each user request into one of five tiers: Quick, Standard, Heavy, Vault, and Learn. Simple edits only take the quick path (about 1.5K tokens), while complex feature development activates the full pipeline. This "tier routing" is its claimed main value, used to reduce unnecessary token overhead in most everyday work.
Main Capabilities
- Tier classification and adaptive execution: The router determines which tier a request belongs to and loads only the content required for that tier.
- Built-in skill library: Comes with 40+ skills, covering common development tasks.
- Subagent prompt system: The agents/ directory contains 19 subagent prompt files, dispatched by Heavy-tier planning. Core agents include: imlazy-planner (planning), imlazy-architect (architecture), imlazy-code-reviewer (code review), imlazy-security-reviewer (security review), imlazy-tdd (test-driven development), imlazy-build-fixer (build fixing), imlazy-e2e (end-to-end testing), imlazy-debugger (debugging), imlazy-docs (documentation), imlazy-refactor (refactoring), imlazy-database (database), imlazy-docs-lookup (documentation lookup). Language reviewers include: imlazy-python-review, imlazy-go-review, imlazy-rust-review, imlazy-java-review, imlazy-cpp-review, imlazy-kotlin-review, imlazy-php-review.
- Obsidian vault integration: Integrated with an Obsidian note vault; specific usage detailsare not elaborated in detail.
How to Use Activate it via the /imlazy slash command, or use the prefix "i am lazy". The router classifies the request and loads only the content for the required tier. The material describes it as self-contained with zero external runtime dependencies, usable by dropping it into any project, with an MIT license valid in perpetuity.
Applicable Scenarios Suitable for users developing with Claude Code, OpenCode, or Codex CLI, especially developers who want to reduce token consumption on everyday simple tasks without sacrificing the engineering workflow for complex features. It is also suitable for team scenarios that require standardized workflows such as code review, security review, TDD, refactoring, and E2E testing.
Limitations and Information Not Yet Provided The material does not provide specific installation commands, configuration steps, detailed determination rules for each tier, the complete list of 40+ skills, the specific integration method for the Obsidian vault, pricing information, platform restrictions, or actual effectiveness data. Regarding testing, the material states that tests verify self-containment (no cross-plugin dependencies), the router's accuracy on fixtures, the token budget of each skill, and the correctness of hook injection. Other details should be based on the repository's actual documentation.
