YouTube shows free Codex+Clawpatch coding stack

A YouTube video posted this week showed a developer workflow built around Codex and Clawpatch, describing the combination as a free setup for AI coding agents. The video, titled “Codex + Clawpatch: This Free Setup Is Really INSANE For AI Coding Agents,” was crawled by search tools on Tuesday and says Clawpatch can scan entire codebases and find bugs. YouTube metadata available through search did not provide a full transcript, so the technical details in the demo could only be verified in part against public documentation. The underlying tools named in the video — Codex, OpenClaw and Clawpatch — all have current public documentation or repositories describing how the stack works. ### Which parts of the demo can be checked against public documentation? Clawpatch’s public site says it is an automated code review tool that “maps a repo into semantic work units,” reviews bounded context and persists findings for audit and follow-up. Its GitHub repository says the tool can “run an explicit fix loop for one finding at a time,” though it also says patching is still behind `clawpatch fix --finding <id>` and requires manual review of the resulting worktree changes. (youtube.com) OpenClaw’s Codex harness documentation says the bundled Codex plugin lets OpenClaw run embedded OpenAI agent turns through the Codex app-server rather than its built-in harness. The same page says Codex can own “low-level agent session” handling, including thread resume, tool continuation, compaction and app-server execution, while OpenClaw keeps approvals, session files and the visible transcript mirror. ### What does “free” appear to mean in this setup? (clawpatch.ai) The YouTube title emphasized “Free,” but the public documentation points to a narrower claim than universally zero-cost software. A LumaDock tutorial published on March 3 says OpenAI supports ChatGPT subscription OAuth in third-party tools like OpenClaw, and says ChatGPT Plus, Pro or Team subscriptions can cover Codex usage inside OpenClaw without per-token API billing. That same guide describes the setup as using an existing ChatGPT subscription rather than a separate API key. (docs.openclaw.ai) OpenClaw’s quickstart documentation also says “most users who want Codex in OpenClaw” should sign in with a ChatGPT or Codex subscription, enable the bundled Codex plugin and use an OpenAI model reference. That means the setup shown in the video may be low-cost for users who already pay for a qualifying ChatGPT plan, but the public materials reviewed do not support saying the entire stack is free in every case. ### How does this stack handle code changes across a project? (lumadock.com) Clawpatch’s documentation says it reviews repositories as “semantic work units” rather than isolated files, with nearby tests and trust boundaries included in the review context. Its GitHub page says findings are stored with enough state for follow-up and that fixes run one finding at a time, which matches the video’s framing around patch-based edits rather than open-ended code generation. (docs.openclaw.ai) OpenAI’s Codex documentation for Windows says the Codex app supports parallel agent threads, worktrees, automations, Git functionality, plugins and skills. The same page says sandboxing is intended to block filesystem writes outside the working folder and prevent network access without explicit approval. ### Did the public sources confirm automated test runs? The YouTube search snippet describes Clawpatch as scanning codebases and finding bugs, but it does not include a transcript line confirming specific test commands. (clawpatch.ai) Clawpatch’s public materials say reviews include nearby tests and validation state, and outside summaries of the project describe a fix loop with audit and validation, but the repository text available in search results stops short of a direct promise that every demo step automatically runs tests in the same way shown on video. (developers.openai.com) The next public checkpoints are likely to come from the same sources already tied to the demo: the YouTube video page for any updated description or comments, Clawpatch’s GitHub repository for CLI changes, and OpenClaw’s Codex harness documentation for setup and runtime requirements, including the current requirement for Codex app-server 0.125.0 or newer. (youtube.com)