CrewAI 1.15.28
CrewAI 1.15.28 introduces the crewai eval command, leverages the LiteLLM catalog for context window management, and delivers multiple bug fixes across deployments, flows, and tool calls.
Why this signal matters
CrewAI release 1.15.28 introduces new features, bug fixes, and documentation updates. Key additions include a new crewai eval CLI command designed to report where a stopped model broke and integration with the LiteLLM catalog to determine context windows. Notable bug fixes span deployment improvements (failure reasons, lockfile checks, ZIP deployment updates via AMP source type reading, and a 60-second timeout for trigger list and payload requests), flow corrections (resuming from human feedback, output type preservation, and flow add-crew package resolution), streamed Anthropic tool call completion events, image input allowances for OpenAI models, UTF-8 BOM JSON decoding in RAG, and preservation of native tool aborts.
Actionable summary
CrewAI v1.15.28 adds a new CLI evaluation command ( crewai eval ), integrates LiteLLM's catalog for context windows, addresses flow human-feedback resume issues, handles streamed Anthropic tool call events, and updates deployment diagnostics.
- Agent usefulness
- 62/100
- Confidence
- 96%
- Canonical data
- JSON + Markdown
What builders should check
- Upgrade CrewAI to version 1.15.28.
- Evaluate the new crewai eval command to inspect stopped model failure points.
- Verify deployment workflows and trigger timeouts if utilizing CrewAI deployment tooling.