What is Codex CLI in Glueprint?
Codex is a first-class native adapter in Glueprint — not a wrapper script, not a re-implementation, and not routed through the generic ACP harness. Four things make it feel native.
Everything below works with the Codex CLI you have today — no migration, no new account for the model, no lock-in. Uninstall Glueprint and your terminal workflow is exactly as it was.
Set up Codex CLI on a host.
Installation is host-specific: repeat these steps anywhere this agent should execute. Portal and mobile can control the session afterward, but the process remains on that host.
Install and authenticate Codex
Install the official package on every host where sessions should run, then complete Codex's own authentication flow.
Detect the CLI in Glueprint
Open Settings -> Agents -> Codex. Leave CLI Path on Auto-detect, choose Detect, and confirm the resolved path and version.
Start and steer a project session
Open a project or worktree, create a Codex session, and send a follow-up while a turn is active to steer it.
npm install -g @openai/codex
codex --versionHow the connection works.
Glueprint adds orchestration and shared surfaces without taking ownership of the coding agent or its provider account.
Glueprint finds the codex binary you already installed, probes the version, and spawns it. Your ChatGPT login or API key, your config, your project's AGENTS.md: all of it comes along, because it's the same CLI you run in the terminal.
The adapter drives Codex through its app-server mode — a long-lived JSON-RPC 2.0 channel rather than one-shot invocations. Prompts, reasoning, tool calls, and diffs arrive as typed events and render as a live timeline on every surface.
Codex's turn-steering is wired through. Send a clarification while the agent is still working — the next tool call uses your new context instead of finishing the wrong plan first.
When Codex wants to write a file or run a command, the request lands as an inline approval card on the desktop, in the browser, or on your phone. Approve, deny, or auto-allow per tool.
Capabilities and discovery.
The status labels separate guaranteed integration behavior from features owned or announced by the installed agent.
| Capability | Status | Behavior |
|---|---|---|
| Structured streaming | Supported | Persistent app-server events render as a typed timeline. |
| Mid-turn steering | Native | Follow-up context is acknowledged while the active turn continues. |
| Permission approvals | Supported | Command and file requests become Glueprint approval cards. |
| Project instructions | Agent | AGENTS.md and Codex configuration load from the working directory. |
| Reasoning effort | Native | Reasoning effort is selectable per session. |
| Explicit interrupt | Supported | Active turns can be interrupted from any connected surface. |
Authentication, ownership, and limits.
Glueprint is the environment around the agent. Credentials, model access, and vendor-specific behavior stay where their owner designed them.
Watch every prompt, tool call, and diff stream in real time — desktop, web portal, iOS and Android. The session runs on your machine; the view follows you.
Permission prompts arrive as approval cards wherever you are. Kick off a refactor at your desk, approve the file writes from the couch.
Don't wait for a wrong turn to finish. Glueprint wires Codex's steering through, so a follow-up message redirects the agent while it's still working.
Run several Codex sessions side by side, each in its own git worktree. Built-in worktree management, diffs, and a three-pane conflict resolver keep them from stepping on each other.
Tile Codex sessions from your laptop, your desktop, and a homelab box into one Command Center mosaic — next to Claude Code and Gemini if you run those too.
Promote Codex to a persistent assistant: persona, constitution, long-term memory, and a daily journal on disk, loaded into every wake from AGENTS.md.
Wake Codex on cron: triage the inbox every 30 minutes, sweep PRs at 9am, run the nightly cleanup. Quiet hours and watermarks prevent re-fires.
Messages, sessions, and output tokens computed from session logs — compare Codex against Claude Code, Gemini, and your ACP agents.
Common questions
Does Glueprint replace the Codex CLI?
No. Glueprint is an Agentic Development Environment that hosts Codex. It spawns the official CLI you already installed and adds the surfaces around it: live session streaming, mid-turn steering, inline approvals, parallel worktrees, scheduling, and remote control.
Do I need an API key or a separate ChatGPT subscription?
No. Glueprint doesn't resell model access. Sessions run through the codex CLI's own authentication — your existing ChatGPT plan or OpenAI API key is exactly what gets used.
Will it use my AGENTS.md and config?
Yes. Sessions launch from your project directory, so AGENTS.md and your Codex configuration load exactly as they do in the terminal.
Can I run Codex from my phone?
Sessions always run on your machines — a laptop, a desktop, or a server with the headless daemon. The iOS and Android app streams them live, lets you send follow-ups, and handles permission approvals from anywhere.
Can I run Codex next to Claude Code?
Yes — that's the point. Each harness is an adapter; sessions from all of them tile into the same Command Center, share the same task boards, and roll up in the same per-harness analytics. See how Glueprint runs Claude Code.
Is my session content private?
Yes. Remote viewing routes AES-256-GCM-encrypted envelopes through the relay, which stores a wrapped key it cannot open — decryption happens on your devices. On the free local tier, session content never leaves your machine at all. See security.