OpenAI lets ChatGPT subscribers use OpenClaw agents – what happens

OpenAI announced that ChatGPT subscribers can now sign in to the open‑source AI agent framework OpenClaw via OAuth, turning their subscription into the billing layer for autonomous agents that run on users’ own hardware [1]. The change means Plus users pay a flat $23 per month to access GPT‑5.4 through OpenClaw, bypassing per‑token API fees [1].

Key takeaways

• OpenAI enabled ChatGPT Plus authentication for OpenClaw, pricing the combined service at $23 per month [1].
• The integration gives OpenAI a direct distribution channel to OpenClaw’s 3.2 million users [1].
• Security researchers have flagged multiple vulnerabilities in older OpenClaw versions, though current releases are patched [1].
• A Tom’s Guide test reported lower API costs but noted usage limits when running an OpenClaw agent with a ChatGPT subscription [2].

OpenAI’s subscription tie‑in with OpenClaw

Sam Altman posted on X that users could “sign in to OpenClaw with your ChatGPT account now and use your subscription there” [1]. OpenClaw, created in November 2025 by Peter Steinberger, quickly became the fastest‑growing GitHub project, amassing 346 000 stars and over three million users [1]. By linking ChatGPT Plus to OpenClaw, OpenAI allows subscribers to log in via OAuth, access the GPT‑5.4 model through the Codex endpoint, and run autonomous agents on their own machines for a flat $23 monthly fee [1]. This contrasts with Anthropic’s April decision to block Claude subscriptions from OpenClaw, a move the source describes as a defensive cost‑control measure [1].

User experience and security backdrop

A Tom’s Guide contributor, Christoph Schwaiger, tried the new setup and found that the ChatGPT‑based subscription saved API expenses compared with per‑token billing, but the service imposed its own limits on agent usage [2]. The broader OpenClaw ecosystem has faced notable security challenges: a remote code execution flaw (CVE‑2026‑25253) and a skills marketplace with hundreds of malicious entries were disclosed earlier in the year [1]. While the latest OpenClaw releases have patched these issues, many installations still run older, vulnerable versions [1].

Why it matters

OpenAI’s move turns a popular open‑source agent platform into a direct channel for its subscription product, potentially converting a sizable portion of OpenClaw’s user base into paying ChatGPT customers [1]. The strategy hinges on the economics of subsidising high‑volume autonomous agent compute with a flat‑rate fee, a gamble that differs from Anthropic’s more restrictive approach [1]. Ongoing security concerns highlight the risk of tying a major brand’s billing and authentication to an ecosystem that has experienced multiple vulnerabilities [1]. Future developments will likely focus on how many OpenClaw users adopt the ChatGPT subscription, how OpenAI manages the compute cost, and whether additional security hardening becomes standard for the platform.