CrewAI

CrewAI Discovery

CrewAI Discovery matches patterns observed over billions of agent runs against your tickets, chats, apps, and workflows. This feature provides a ranked list of automation opportunities based on effort, value, and readiness, helping users identify what to automate before they start building.

Agentic use case generator

Powered by billions of agent runs, this feature generates use cases that are relevant to your specific needs. It helps users visualize potential automation opportunities and aligns team efforts towards common goals.

Interactive suggestions

This feature refines and improves recommended automations through interactive suggestions. Users can engage with the suggestions to better tailor automation opportunities to their specific workflows.

Shareable presentation format

Agent automations can be presented in a shareable format, which accelerates team alignment. This feature ensures that all stakeholders are on the same page regarding automation strategies.

One-click context

This feature provides an accelerated path to build automations by offering one-click context. It simplifies the process of understanding the necessary context for automation tasks.

No-code visual editor

The no-code visual editor allows users to build multi-agent workflows without any programming knowledge. It is designed for ease of use and can be exported to Python for further customization.

Code-first API

For users who require more control, the code-first API is built to provide total control over agent orchestration. This feature is ideal for developers looking to implement complex workflows.

Role-based agents

Role-based agents help separate and simplify agent orchestration. This feature allows users to define roles for different agents, making it easier to manage complex workflows.

Deterministic agent workflows

Users can create deterministic agent workflows that ensure predictable outcomes. This feature is crucial for maintaining consistency and reliability in automated processes.

Control Plane

CrewAI's Control Plane sits in the execution path of every workflow, ensuring that every agent interaction is observable, compliant, and reversible. This feature provides a layer of oversight for all automated processes.

Real-time tracing

This feature allows for real-time tracing of every LLM call, tool call, and memory read, providing full cost accounting. Users can monitor the performance and costs associated with their workflows.

RBAC and audit

Role-Based Access Control (RBAC) and audit features provide granular control and immutable audit trails. This ensures that all actions taken within the platform are logged and can be reviewed for compliance.

Human-in-the-loop approval gates

This feature allows for human intervention during execution, providing approval gates that ensure critical decisions are overseen by a human operator. It enhances the reliability of automated processes.

Runtime hooks

Runtime hooks can inject PII redaction and policy checks at every LLM and tool call. This feature ensures that sensitive information is handled appropriately throughout the automation process.

Automated and human-guided training

CrewAI turns every production run into training data, allowing for continuous improvement of agent performance. This feature combines automated processes with human oversight to enhance accuracy.

Multi-LLM testing

This feature allows for model swapping at runtime, enabling users to find the right model for every workflow. It enhances flexibility and performance by allowing users to test different models in real-time.

Evaluation for native tracking

The evaluation feature provides native tracking with expanded sophistication powered by tools like Arize, Galileo, DataDog, and Patronus. This allows users to gain deeper insights into their workflows.

Real-time tracing for every LLM call

Similar to the governance feature, this allows for real-time tracing of every LLM call, tool call, and memory read, ensuring that users can monitor their workflows effectively.