CrewAI 1.15.34
CrewAI v1.15.34 adds a skill search command, CLI deployment enhancements, flow input passing, improved JSON output refactoring, and documents the removal of the 1.15.29 agent executor.
Why this signal matters
CrewAI version 1.15.34 brings a series of features, bug fixes, and breaking changes. Key features include the addition of a standalone LiteLLM catalog, the crewai skill search command to discover organizational skills, support for passing --inputs into Python flows during crewai run , and deploy create --zip support that logs source commit info. Documentation was updated to detail agent planning using planningconfig and to migrate the reasoning guide. Key bug fixes enhance reliability: empty LLM responses are retried once before raising a typed error, malformed JSON matching output models is refactored rather than hard-failing, and context window limits are strictly respected. LiteLLM streaming exceptions and token usage reporting for tool/structured calls have been resolved. The release notes also document breaking changes concerning the removal of the 1.15.29 agent executor, accompanied by migration notes.
Actionable summary
CrewAI 1.15.34 introduces features like crewai skill search , deploy create --zip , and passing inputs via crewai run --inputs to Python flows. It includes robust LLM error handling (empty answer retries, repeated tool call handling, context window adherence) and records the removal of the 1.15.29 agent executor with migration guidance.
- Agent usefulness
- 79/100
- Confidence
- 96%
- Canonical data
- JSON + Markdown
What builders should check
- Review migration notes if relying on the legacy 1.15.29 agent executor.
- Verify compatibility with raised dependency floors for joserfc, pyasn1, and pypdf.
- Check configurations for OpenAI-compatible providers to ensure non-placeholder API keys are supplied to avoid MissingAPIKeyError .