OpenAI Codex Now Reads and Writes Across Multiple Project Folders at Once
Codex local projects can now span multiple folders, letting the agent read and write across them while keeping one Git root in charge.
- Multi-folder projects: Codex local projects can now include code, docs, and reference files from multiple directories at once.
- Primary folder rules: One folder stays the Git root and handles PRs, review,
AGENTS.md, and skill discovery; others are readable/writable by the agent. - Desktop app only: Configured via the ChatGPT desktop app settings pane; CLI already had
--add-dirsupport. - Config is shareable: The setup is stored in
.codex/at the project root and can be checked into Git for team use. - Key use cases: Split frontend/backend repos, shared libraries, documentation vaults, and cross-project refactoring.
- Availability: Included in existing ChatGPT Pro ($200/mo), Team ($25/user/mo), and Enterprise plans.
OpenAI Codex shipped a quality-of-life update many users had been requesting: local projects can now include files from multiple folders at once. Previously, a Codex project was locked to a single directory, so the agent could only read and write within that one folder. Now you can bring in related code, docs, and reference files from sibling or parent directories into a single project view.
One project, many roots
The Codex app previously allowed selecting only one project folder per workspace, which created real friction for setups where source code lives in one folder and documentation or shared libraries live in another. A common example: an AI workspace for current deliverables paired with a knowledge vault as a sibling folder. The vault might be a long-term reference base shared with other tools, but Codex had no way to read it while writing output to the workspace.