> AI Coding Agent Guides
Real terminal workflows for bounded coding Loops, your own model API key, and read-only handoff from Codex, Claude Code, Cursor, Kimi Code, or Grok.
- GUIDE_015 minREAD
Running an AI Coding Agent in a Repository With No Tests
No executable checks does not mean no result. It means the Loop records a versioned decision you have to acknowledge, and a Gate that cannot lean on evidence it never had.
- GUIDE_026 minREAD
PASS, NEEDS_WORK, ERROR: Reading an AI Coding Agent Result
Most coding agents end in prose. A Loop ends in one of a small set of results, each with a different evidential weight and a different next action.
- GUIDE_036 minREAD
Run an AI Coding Agent Until Checks Pass—or It Stops Clearly
Turn a one-shot coding prompt into a bounded write, check, review, and fix Loop with an explicit Gate or lifecycle result.
- GUIDE_045 minREAD
BYOK AI Coding Agent CLI: Use Your Own Model API Key
Connect your own model provider, keep Wukong orchestration local, and understand the real privacy and cost boundary.
- GUIDE_056 minREAD
Cross-Agent Coding Session Handoff: Codex, Claude Code, Cursor, Kimi Code, and Grok
Switch coding agents without replaying old tools, inheriting permissions, or trusting stale test results from the source session.
Ready to run the Loop?
Install the free CLI, connect your own model provider, and run the workflow in a real repository.