Cursor Cloud Agents
Connect AI SmartTalk with Cursor Cloud Agents to launch autonomous AI coding assistants directly from your conversations or workflows. Cursor Cloud Agents can read repositories, generate code, create commits, and open pull requests automatically.
Overview
The Cursor Cloud Agents integration enables you to:
- Launch AI coding agents on your GitHub repositories
- Automate code generation through natural language instructions
- Create pull requests automatically when tasks are complete
- Monitor agent progress and view conversation history
- Add follow-up instructions to running agents
- Trigger coding tasks from SmartFlow workflows
Prerequisites
Before you begin, ensure you have:
- An active AI SmartTalk account
- A Cursor account with Cloud Agents access
- A Cursor API key (obtained from your Cursor dashboard)
- GitHub repositories accessible from your Cursor account
Step-by-Step Setup
Step 1: Get Your Cursor API Key
- Go to your Cursor Cloud Agent dashboard
- Navigate to your API settings
- Generate a new API key for AI SmartTalk
- Copy the API key
Step 2: Access Cursor Integration
- Log into your AI SmartTalk account
- Navigate to Administration → My Assistants → [Your Assistant Name]
- Click on Connectors in the left sidebar
- Locate Cursor Cloud Agents and click on it
Step 3: Configure the Connection
- Click Setup Configuration
- Enter your Cursor API Key
- Click Save Configuration
- AI SmartTalk validates your API key automatically
Step 4: Sync Repositories
Once connected, AI SmartTalk automatically syncs accessible GitHub repositories from your Cursor account. You can also manually refresh the repository list at any time.
Dashboard
Once connected, the Cursor integration dashboard provides a complete overview:
Stats Overview
| Metric | Description |
|---|---|
| Total Agents | Number of agents launched |
| Running | Currently active agents |
| Finished | Completed agents |
| Available Models | AI models available for agents |
Recent Assistants
The dashboard lists your most recent agents with:
- Agent ID for reference
- Status: Running, Completed, Failed, Creating, or Cancelled
- Repository the agent is working on
- Creation date
- Quick actions: View Details and View PR
The dashboard auto-refreshes every 30 seconds when agents are running.
Launching an Agent
From the Dashboard
- Click Launch New Assistant
- Fill in the launch form:
| Field | Required | Description |
|---|---|---|
| Repository | Yes | Select a GitHub repository from the dropdown |
| Task Description | Yes | Describe what the agent should do in natural language |
| Branch | No | Git branch to work on (default: main) |
| AI Model | No | Select a specific AI model for the agent |
| Auto-create PR | No | Automatically create a pull request when done (default: enabled) |
| Branch Name | No | Custom branch name for the pull request |
- Click Launch Agent
Task Description Examples
- "Add unit tests for the authentication module"
- "Refactor the database connection pool to use connection recycling"
- "Fix the bug where user sessions expire prematurely on mobile"
- "Add TypeScript type definitions for all API response objects"
- "Create a REST API endpoint for user preferences"
Managing Agents
View Agent Details
Click View Details on any agent to see:
Overview Tab:
- Agent ID and current status
- Repository and branch information
- Pull request link (if created)
- Summary of work performed
- Creation timestamp
Conversation Tab:
- Full message history of the agent's work
- User instructions and AI responses
- Step-by-step progress log
Add Follow-Up Instructions
You can send additional instructions to a running agent:
- Open the agent details
- Enter your follow-up instruction (e.g., "Also add error handling for the edge case when the user is not found")
- The agent continues its work with the new context
Delete an Agent
To clean up resources:
- Click the delete action on the agent card
- Confirm the deletion
- The agent and its resources are removed
SmartFlow Integration
Cursor Cloud Agents can be used as actions in SmartFlow workflows, enabling powerful automation scenarios.
Available SmartFlow Actions
| Action | Description |
|---|---|
| Launch Coding Agent | Start an agent on a repository with specific instructions |
| Add Followup | Send additional instructions to a running agent |
| Get Agent Status | Check the progress and status of an agent |
| Delete Agent | Remove an agent and clean up resources |
Example: Auto-Fix on Bug Report
Trigger: Chat Service (when user reports a bug)
Actions:
1. AI Request:
Prompt: Extract the bug description and affected repository
2. Cursor Launch Agent:
Repository: {{extracted_repository}}
Task: "Fix this bug: {{bug_description}}"
Auto-create PR: true
3. Send Message:
Text: "I've launched a coding agent to fix this. I'll notify you when the PR is ready."
4. Wait / Poll:
Check Agent Status periodically
5. Send Message:
Text: "The fix is ready! Review the PR here: {{pr_url}}"
Example: Scheduled Code Improvements
Trigger: Scheduled Workflow (weekly)
Actions:
1. Cursor Launch Agent:
Repository: my-org/my-app
Task: "Review and update all deprecated dependencies"
Branch: chore/update-deps
Auto-create PR: true
For detailed SmartFlow action parameters, see Cursor SmartFlow Actions.
Use Cases
Developer Support Bot
Let users request code changes through chat:
- "Can you add input validation to the signup form?"
- "Please create a migration script for the new user fields"
Automated Bug Fixing
Trigger coding agents from bug reports or monitoring alerts to auto-generate fix PRs.
Code Review Automation
Launch agents to review code, add tests, or improve documentation based on PR feedback.
Scheduled Maintenance
Set up recurring workflows to update dependencies, run linting fixes, or generate reports.
Troubleshooting
| Issue | Solution |
|---|---|
| "Invalid API key" | Regenerate your API key from the Cursor dashboard |
| "Repository not found" | Click Refresh to re-sync repositories. Ensure the repo is accessible from your Cursor account |
| Agent stuck in "Creating" | The agent may be queued. Wait a few minutes and check status again |
| Agent failed | Open agent details to view the conversation and error messages |
| PR not created | Check that auto-create PR was enabled and the agent completed successfully |
| No repositories available | Ensure your Cursor account has GitHub repositories connected |
Managing the Integration
| Action | How |
|---|---|
| Connect | Enter API key in the integration configuration |
| View dashboard | Open the Cursor integration page |
| Sync repositories | Click Refresh to update the repository list |
| Disconnect | Remove the API key from the configuration |
Best Practices
- Write clear task descriptions: The more specific your instructions, the better the agent's output
- Use appropriate branches: Don't launch agents directly on
main— let them create PRs for review - Enable auto-create PR: Makes it easy to review and merge agent work
- Monitor running agents: Check the dashboard regularly or set up SmartFlow notifications
- Review before merging: Always review agent-generated PRs before merging to production
- Start small: Begin with well-scoped tasks before attempting large refactors