crewAI 1.15.17 adds declarative conversational flows to simplify multi-agent orchestration
Define AI agent workflows conversationally without writing complex orchestration code
crewAI Inc has rolled out version 1.15.17 of its popular multi-agent orchestration framework, introducing declarative conversational flows as the headline feature. The update allows developers to define conversational interactions between AI agents declaratively, rather than hard-coding the back-and-forth logic. This is powered by new built-in conversational methods and a clearer opt-in mechanism, making the conversational mode unmistakable for users. The team also added support for handling oversized single messages through automatic chunking, which prevents context-window overflows in long-running agent conversations.
Beyond new features, this release packs several important bug fixes. SSRF (server-side request forgery) checks are now pinned to each redirect hop and peer IP, improving security when agents interact with external URLs. Tool errors are now accurately attributed to the specific tool that failed, simplifying debugging. The release also resolves issues with native tool calls over OpenAI's Responses API, ensuring smoother integration. Documentation was updated with a snapshot and changelog for v1.15.16. With 57.4k GitHub stars and 8.2k forks, crewAI continues to strengthen its position as a go-to framework for production-ready agent workflows, especially for teams building complex, conversational AI systems.
- Declarative conversational flows with built-in conversational methods for simpler agent orchestration
- Security fix: SSRF checks pinned per redirect hop and peer IP
- Fixed native tool calls over OpenAI Responses API and oversized message chunking
Why It Matters
For AI engineers, declarative conversational flows slash boilerplate, enabling faster, more secure multi-agent systems in production.