Research
Use /research_codebase to explore and document how existing code works. AI agents write findings to the thoughts directory so knowledge persists across sessions.
Hyprlayer CLI
AI agents research your codebase, build implementation plans, execute them phase by phase, and validate the results against a thoughts directory your whole team shares.
terminal
$ brew install brightblock/tap/hyprlayer
$ hyprlayer thoughts init
$ hyprlayer ai status
Status: installed
Version: 1.6.2
Claude Code: ~/.claude/
Codex: ~/.codex/
# Now run /research_codebase in Claude Code or CodexHow It Works
Section titled “How It Works”Hyprlayer installs a CLI (hyprlayer) plus 14 skills and 19 sub-agents for Claude Code and Codex. The CLI manages a shared thoughts repository, so your agents keep their context on the codebase between sessions and across your team.
Plan
Use /create_plan to build phased implementation plans with success criteria. You and the agent research and iterate on the plan before anyone writes code.
Implement
Use /implement_plan to execute plans phase-by-phase with verification at each step. The agent follows the plan’s intent while adapting to what it finds.
Validate
Use /validate_plan to verify the implementation against the plan’s success criteria and identify any deviations or issues.
Works with your AI
Section titled “Works with your AI”Hyprlayer installs both bundles the first time you run it. There is no tool to pick. OpenCode reuses them: Claude models use the Claude bundle, every other model uses the Codex bundle.