workflow-convergence-verifier

workflow-convergence-verifier MCP Connector for Claude

A+

Analyzes agent workflows to detect stability, convergence, or infinite loops.

3 tools Official Updated Oct 1, 2026 Official Vinkius Partner

This MCP server provides diagnostic tools to monitor agent execution. It helps identify if a multi-step workflow has stabilized using verify_convergence, detects repetitive loops with detect_oscillations, and measures structural changes via analyze_state_stability. It is designed to solve the infinite loop problem in complex agent pipelines by analyzing state similarity and hash patterns.

convergenceoscillationstabilityworkflow-analysisagent-debugging

3 tools expose this connector's capabilities to your AI agent.

analyze_state_stability

Provides a granular breakdown of how much the agent's internal structure is changing over time

detect_oscillations

Identifies if the agent is trapped in a repetitive loop (cycling through the same states)

verify_convergence

Determines if the agent's workflow has reached a stable, finished state

See how to talk to your AI agent using workflow-convergence-verifier.

Check if my current agent workflow has finished or is stuck.

The workflow has converged at step 5, indicating the agent has reached a stable state.

Is the agent stuck in a loop?

Yes, oscillation detected. The agent is cycling through 3 repetitive states.

Analyze the stability of the last 10 steps of the workflow.

The average similarity is 0.98 with a low volatility score, suggesting high stability.

The `verify_convergence` tool checks if the similarity between consecutive state key sets remains above a defined threshold for a specific number of steps.

Related Connectors