Does /goal work in the EU?

Unverified. No explicit evidence found. The provided pages cover the third-party Loop Library and say nothing about regional availability of /goal in Codex. Last verified 14 Nov 2026.

How much does /goal cost?

Pricing unknown. Feature of OpenAI Codex; no separate pricing found in the provided material. The Loop Library prompts are free to copy.

Where was /goal shown?

In 4 creator videos by Matthew Berman. Most recently by Matthew Berman on 5 Aug 2026, at 11:31 in "Master Codex with these 15 Tips".

/goal

by OpenAI

Website

Codex command that keeps an agent working on a long-running goal until a verifiable or model-judged condition is met.

UnverifiedAgents & Automation CodingApp

Updated 14 Nov 2026EU status checked 14 Nov 2026Last shown 5 Aug 2026

/goal at a glance

What is /goal?: /goal is an app from OpenAI. Codex command that keeps an agent working on a long-running goal until a verifiable or model-judged condition is met.
Who makes /goal?: /goal is made by OpenAI.
Does /goal work in the EU?: Unverified. No explicit evidence found. The provided pages cover the third-party Loop Library and say nothing about regional availability of /goal in Codex. Last verified 14 Nov 2026.
How much does /goal cost?: Pricing unknown. Feature of OpenAI Codex; no separate pricing found in the provided material. The Loop Library prompts are free to copy.
Where was /goal shown?: In 4 creator videos by Matthew Berman. Most recently by Matthew Berman on 5 Aug 2026, at 11:31 in “Master Codex with these 15 Tips”.

Sources: the vendor's own pages (linked above) and creator videos at the linked timestamps. EU status is information, not legal advice; see how we check.

About

/goal (also called goals or loops) is a command in OpenAI's Codex that keeps a coding agent iterating toward an overarching objective instead of stopping after a single response. The agent continues until a verifiable condition or an LLM-judged goal is reached. Limits such as a time cap can be set.

According to creator Matthew Berman, goals can keep an agent running for hours or days, for example a benchmark loop that iterates until a score reaches 90%. The linked Forward Future Loop Library is a third-party collection of 85 copyable loop prompts (docs sweeps, test coverage, accessibility repair, dependency triage, source checks and others), each with a stated stopping condition. It also offers a "Loopy" agent skill for finding, writing and running loops.

The page material does not document /goal's official syntax, pricing or availability. It is accessed from within Codex. Long runs depend on well-defined, checkable stopping conditions, and loops without a clear check or limit may run indefinitely.

Key features

Runs the agent repeatedly until a goal is met
Supports verifiable or LLM-judged completion conditions
Optional limits such as a time cap
Can run for hours or days on benchmark-style tasks
Community Loop Library of 85 reusable loop prompts

4 moments

As seen on YouTube

Watch the moment: the exact second top creators show it. Videos load only when you press play, via youtube-nocookie.com.

Matthew Berman5 Aug 2026

Live demo

Master Codex with these 15 Tips

Command that makes the agent run for a long time until a verifiable or LLM-judged goal is reached. Limits such as a time cap can be set. The creator has had agents running for days with it.

11:31Open on YouTube

Matthew Berman14 Jul 2026

You aren't using Codex like me...

Loops, called goals in Codex, keep the agent working toward an overarching goal for hours or days. The creator runs a benchmark loop that iterates until the score reaches 90%.

9:12Open on YouTube

Matthew Berman19 Jun 2026

Live demo

7 INSANE loops you need to try right now

A Codex feature that makes the agent keep working until a stated condition is met. Demoed with the sub-50ms page-load loop, which ran nearly 50 minutes. The creator notes Claude Code has a similar /goal feature.

2:55Open on YouTube

Matthew Berman9 Jun 2026

Live demo

Only the best are using them...

Coding technique where you design a trigger and a verifiable goal and let agents prompt themselves until the goal is met, rather than prompting manually. Caveats: hard to set up for non-deterministic goals and expensive in tokens.

0:00Open on YouTube