Bug Report: Cursor AI Agent Stops Responding After First Message
Clear Description of the Bug
Primary Issue: The Cursor AI agent completely stops responding after the first message in any conversation. When attempting to send a second message, the agent shows āgeneratingā status but never produces any output, making it impossible to have any meaningful interaction with the AI. This issue occurs across all AI models available in Cursor.
Secondary Issue: When the agent does reach the 25 tool call limit, clicking the āresume the conversationā button also fails - the agent appears to start generating a response but never completes it.
How to Reproduce the Bug
Primary Issue (Most Critical):
- Start any conversation with the Cursor AI agent
- Send any initial message and wait for the agent to respond
- Attempt to send a second message (any message)
- Observe that the agent shows āgeneratingā status but never produces any output
- Notice that no further communication with the agent is possible
Secondary Issue (25 Tool Call Resume):
- Start a conversation that reaches the 25 tool call limit
- Click the āresume the conversationā button when prompted
- Observe that the agent shows āgeneratingā status but never actually resumes
- Notice that typing ācontinueā or other commands also fails to restart the agent
System Information
- Cursor Version: 51.1 (recently updated)
- Operating System: macOS Sequoia 15.2
Impact on Cursor Usage
This bug completely breaks all AI functionality in Cursor. The issue:
- Prevents any multi-turn conversation with the AI agent
- Makes it impossible to ask follow-up questions or clarifications
- Blocks all iterative development workflows that rely on AI assistance
- Renders the AI assistant feature completely unusable for any practical purpose
- Affects even the most basic interactions, not just complex tasks
Additional Context
- Issue started occurring yesterday
- Primary problem: Cannot send any second message to the agent, regardless of complexity
- Secondary problem: Resume function after 25 tool calls also fails
- Affects all AI models: Issue occurs with Claude Sonnet 4, Claude 3.5, and Gemini Pro 2.5
- Critical finding: Existing conversations from a couple days ago work perfectly fine when resumed - this issue only affects newly created conversations
- Problem persists after complete reinstall of Cursor application
- Agent appears to initialize response generation but fails to complete it
- This affects the most basic AI interactions, making the feature completely unusable for new chats
Troubleshooting Steps Already Attempted
Updated Cursor to version 51.1
Completely uninstalled and reinstalled Cursor
Tested multiple AI models (Claude Sonnet 4, Claude 3.5, Gemini Pro 2.5)
Attempted using ācontinueā command for the resume issue
Tried various types of messages for the second message issue
None of these steps resolved either the primary or secondary issue
This appears to be a critical regression that completely breaks the core AI functionality in Cursor.