[FEATURE] Integration of LoopHalter for automatic infinite loop detection.

Feature Area

Core functionality

Is your feature request related to a an existing bug? Please link it here.

No, this feature request is not related to an existing bug report. It is a proactive enhancement to prevent infinite loop scenarios and optimize API cost management.

Describe the solution you'd like

Proposer

Jacopos311

Target Component

Core Execution Flow / Multi-Agent Coordination

Problem Description

During extended autonomous agent sessions, agents can occasionally fall into infinite communication loops or validation deadlocks (e.g., Agent A constantly requesting modifications and Agent B returning the same response pattern). In production or heavy testing environments, this behavior leads to rapid API token exhaustion and unexpected costs before manual intervention occurs.

Proposed Solution

I have developed LoopHalter, a lightweight, zero-dependency Python middleware designed to intercept agent outputs, monitor token consumption, and break execution cycles when semantic or repetitive loops are detected.

Since it is built entirely using the Python Standard Library, it introduces minimal overhead and can be easily integrated into existing pipelines.

I would like to propose integrating this logic or offering it as a built-in feature/middleware option within CrewAI to provide native budget and loop safety guards for users.

Documented Implementation & Codebase

• GitHub Repository: https://github.com/Jacopos311/LoopHalter.git
• PyPI Package: https://pypi.org/project/LoopHalter/

I am open to feedback from the core team on how to best structure a Pull Request or adapt the sequence matching logic to align with CrewAI's architecture.

Describe alternatives you've considered

No response

Additional context

No response

Willingness to Contribute

Yes, I'd be happy to submit a pull request

Start by reviewing the linked LoopHalter repository and PyPI package, then inspect CrewAI’s core execution flow and multi-agent coordination boundaries. Determine where loop detection, token monitoring, and execution stopping would fit, and clarify the expected behavior and acceptance criteria with the maintainers before implementation. Done requires an agreed integration design and a reviewable contribution scope.