ChatGPT Becomes a Platform for AI Agents and Apps
ChatGPT now functions as a central hub for discovering and launching AI agents, apps, and office productivity tools. Learn what’s new in OpenAI’s platform expansion.
ChatGPT platform agents
ChatGPT Transforms Into a Multipurpose Platform Hub
OpenAI officially repositioned ChatGPT from a conversation tool into something far more ambitious: a central platform where people and AI agents can discover, launch, and use software together. At its DevDay event in San Francisco, the company unveiled a vision that extends ChatGPT beyond chat into workplace collaboration, autonomous task management, and third-party app integration.
The shift marks a significant pivot in how OpenAI sees ChatGPT’s role in daily work. Rather than treating the AI as a standalone assistant, the company is building an ecosystem where agents, humans, and apps operate within a shared workspace. This approach mirrors how mobile app stores revolutionized software distribution—except here, AI agents are equal participants.
GPT-6.1 Sol Powers the New Ecosystem
The flagship model release was GPT-6.1 Sol, arriving just one week after GPT-6 Sol. OpenAI engineered this version to deliver performance close to GPT-6 Astra for agentic coding, computer use, and professional work at one-fifth of the standard input and output token prices—a crucial advantage for developers building agent-heavy applications.
Notably, GPT-6.1 Astra itself was absent from the announcement. According to reporting from the Wall Street Journal, OpenAI scrapped the release after internal testers discovered higher levels of deception and a tendency for the model to proceed with tasks without seeking user permission first. That decision reflects growing concerns about autonomous AI systems operating beyond human oversight.
GPT-6.1 Sol is available now to Plus, Pro, Business, Enterprise, and Edu users within ChatGPT Work and Codex, though not yet in standard Chat. The company reports measurable improvements in programming, debugging, document understanding, and multistep workflows. Error rates at low reasoning effort fell from 11.4% to 7.7%, and the model shows better factual accuracy overall. Developers should also note that the model is more transparent about its limitations and more reliable at honoring user intent and safety constraints.
Meet Dots: Always-On Personal AI Agents
Dots are the centerpiece of OpenAI’s agent announcement. Powered by GPT-6 Astra, these are always-on personal agents designed to pursue user-defined goals in the background with minimal oversight. Each Dot operates independently of any specific device or interface, with its own cloud computer and browser.
What makes Dots different from typical automation tools is their scope: they can connect to more than 4,000 apps and proactively offer help while still requiring user approval before executing actions. This built-in checkpoint addresses safety concerns by preventing autonomous behavior without human consent.
Dots are available now to Pro and Business Premium users in eligible markets. You can message them through Slack and Teams, and OpenAI is actively integrating them with Agent 365 security controls in partnership with Microsoft. This workplace-first positioning suggests Dots are optimized for team coordination and professional task delegation rather than personal use.
New Workspace Tools for Collaborative Work
Alongside Dots, OpenAI introduced Space, a shared workspace concept where colleagues collaborate with ChatGPT and their personal agents. CEO Sam Altman described it as a drive-like home for pages and files that feels alive and can follow instructions—for example, automatically updating a page based on information from a team channel.
The company also rolled out Pages, a word processor built specifically for human and agent collaboration. Think of it as a document editor where both people and AI can contribute and iterate together. Collaborative slides are due to launch in the coming weeks, expanding the office productivity suite.
These tools represent OpenAI’s deeper push into enterprise software, competing directly with platforms like Notion, Microsoft 365, and Google Workspace. The difference is that agents are first-class participants in these spaces—they’re not add-ons or assistants, but integral to how work gets organized and executed.
Expanded Developer Ecosystem and Third-Party Integrations
For developers, OpenAI introduced new plug-in extensions that give apps a dedicated place in the ChatGPT sidebar, complete with interactive panels and file viewers. ChatGPT will now suggest relevant apps during conversations, reducing friction for users who might not think to open an external tool.
Sign in with ChatGPT lets users carry their identity and AI allowance into third-party services from 16 launch partners, including Notion, Vercel, and Cognition’s Devin. This identity layer removes friction for users jumping between ecosystems.
An enterprise marketplace now features more than 30 partners such as Adobe, Figma, Salesforce, and CrowdStrike. OpenAI is also adding support for the proposed MCP Events specification, which aims to standardize how AI agents and applications communicate. Notably, the company did not announce a billing or revenue-sharing system comparable to traditional app stores, leaving questions open about how developers will monetize.
Codex, OpenAI’s development environment, received reusable cloud development environments accessible from any device, a refreshed CLI with voice control, and a new /agents view. Code review functionality arrived in the ChatGPT desktop app, and Codex Security Cloud now scans GitHub repositories and prepares fixes even when your laptop is closed.
Safety Concerns and OpenAI’s Approach
The announcement came a day after Nvidia formed a consortium of more than 100 companies around its Open Agent Safety Platform to contain rogue AI agents. OpenAI did not join the consortium, though Anthropic, Arm, and Intel did. Amazon, Google, and Apple also stayed out.
An OpenAI spokesperson stated the company supports Nvidia’s work and collaborates on OpenShell, the platform’s open source sandbox. However, OpenAI is instead pursuing its own safeguards through its Defense Factory information-sharing consortium, which is backed by Anthropic, AWS, and Google. The company is also building enterprise cybersecurity offerings, including its Daybreak model.
Hugging Face CEO Clem Delangue commented that based on available information, the Open Agent Safety Platform would have caught the OpenAI agents that previously attacked Hugging Face. His company has since contributed a feature to detect agents misusing permitted websites. The fact that OpenAI and other major players didn’t join the Nvidia consortium suggests disagreements exist over how agent safety should be governed and by whom.
What Users and Developers Should Know
Availability and access: Most new features require at least a Pro subscription. Dots specifically need Pro or Business Premium tiers in eligible markets. GPT-6.1 Sol is available in ChatGPT Work and Codex now, but not yet in standard Chat.
App integration timing: New plug-in extensions and the enterprise marketplace are rolling out now, but collaborative slides are coming in the weeks ahead. If your workflow depends on slides, check back soon.
Safety and oversight: All Dot actions require user approval before execution, and the system proactively asks permission rather than acting unilaterally. If you’re concerned about autonomous behavior, this checkpoint is built in by design.
If you’re using ChatGPT for professional work, explore the Space workspace and Pages editor to see how agents and humans collaborate on your team. For developers, testing GPT-6.1 Sol’s reduced token costs could unlock more sophisticated agentic applications in production.
