We’re launching Lyria 3.5 in Google Flow Music, with advances across musicality, lyrics, vocals, and creative control
deepmind.google·6d ago
AI Engineering
OpenAI Blog·Apr 23·5 min read
Plugins and skills
TL;DR
Large language models like Codex struggle to interact with external tools and data sources without manual integration. OpenAI built plugins and skills—modular extensions that let Codex connect to APIs, databases, and workflows automatically.
✦ Why It Matters
Engineers can now build AI systems that autonomously integrate external tools, data, and workflows to automate complex tasks reliably.
Key Takeaways
- Define plugins to expose APIs and data sources as callable functions for Codex.
- Encapsulate multi-step processes as skills to enable reusable workflow composition.
- Use semantic matching to automatically select relevant plugins and skills for tasks.
- Chain plugins and skills together to automate end-to-end workflows without manual intervention.
Full Summary
Codex, a code-generating AI model, excels at writing software but lacks native ability to access external tools, APIs, or real-time data needed for practical automation. OpenAI developed plugins and skills—reusable software modules that act as bridges between Codex and external systems.
Plugins enable Codex to call APIs and retrieve live data; skills encapsulate repeatable workflows as callable functions. The approach uses semantic matching to determine which plugins or skills apply to a given task, then chains them together in execution order.
This architecture allows developers to define custom integrations once and reuse them across multiple automation tasks. Results show improved task success rates and reduced manual intervention when automating multi-step workflows that require external data or tool access.