AI Integration
RESTK connects directly to your AI coding assistant — Claude Code, Claude Desktop, or Cursor. Browse collections, run requests, generate tests, debug failures, and audit APIs using natural language.
RESTK connects directly to your AI coding assistant — Claude Code, Claude Desktop, or Cursor. Talk to your APIs using natural language.
Getting Started
One-Click Setup
Open RESTK
Launch the RESTK app on your machine.
Go to Settings → AI Integration
Select your AI assistant from the list (or use the profile menu shortcut).
Click Setup
RESTK handles the configuration automatically. Start a new session in your AI assistant to connect.
Manual Setup
Claude Code:
claude mcp add --transport stdio --scope user restk -- /Applications/Restk.app/Contents/Resources/restk-bridge
Claude Desktop — add to ~/Library/Application Support/Claude/claude_desktop_config.json:
{
"mcpServers": {
"restk": {
"command": "/Applications/Restk.app/Contents/Resources/restk-bridge"
}
}
}
Cursor — add to ~/.cursor/mcp.json:
{
"mcpServers": {
"restk": {
"url": "http://localhost:52452/mcp"
}
}
}
Prerequisites
- RESTK app must be running
- Collections must have Share with AI enabled
- Credentials and secrets are never exposed to AI
For detailed configuration and troubleshooting, see the MCP Setup Guide.
Browsing Your API Workspace
Ask your AI assistant to read your API data. Just say what you want to see.
"Show me all my collections"
"List the requests in my Users folder"
"Show me the details of the Login request"
"What environments do I have?"
"Show me the active environment variables"
"Show me the last 10 API calls I made"
"Search my history for requests that hit the /payments endpoint"
"Show performance trends for the Create Order request"
"Show me the Nova scripting API reference"
Script Reference First
Always ask the AI to read the scripting reference before writing test scripts — it ensures the correct syntax.
Running & Testing APIs
Execute a Request
"Run the Login request"
RESTK will show a confirmation dialog in the app. The request runs with full variable substitution, authentication, and scripts.
"Run the Create User request with a 60 second timeout"
Execute a cURL Command
"Run this cURL command: curl -X POST https://api.example.com/users -H 'Content-Type: application/json' -d 'the JSON body'"
Execute any HTTP request directly — always requires your approval.
Re-run with Changes
"Re-run the Get Users request but change the page to 5 and limit to 100"
Export as cURL
"Export the Create Order request as a cURL command I can paste in my terminal"
Creating & Organizing APIs
Create a Collection
"Create a new collection called Payment Gateway API"
Create Folders
"Create a Users folder and an Orders folder in my Payment Gateway collection"
Create Requests
"Create a POST request called 'Create User' in the Users folder. URL is {{baseUrl}}/api/users, with a JSON body containing name, email, and role fields"
"Create a GET request 'List Orders' at {{baseUrl}}/api/orders with page and limit query parameters"
Update Requests
"Add authentication headers to the Create User request"
"Change the Login request method from GET to POST"
"Add a test script to Create User that verifies the status is 201 and captures the user ID"
Set Up Environments
"Create a Development environment with baseUrl set to http://localhost:3000 and apiVersion set to v1"
Set Up Authentication
"Set Bearer token authentication on the Payment Gateway collection using the authToken variable"
Import a Collection
"Import my Postman collection from the exported JSON file"
Analyzing & Debugging
Performance Analysis
"How is the Create Order endpoint performing? Show me stats for the last 30 days"
Returns mean, median, P95, P99 response times, error rates, and whether performance is improving or degrading.
"Show me a timeline of all error responses from this week"
Response Analysis
"Analyze the latest response from the Get Products endpoint"
"What's the JSON schema of the Get Users response?"
Compare Responses
"Compare the last two responses from Get Users — what changed?"
Find Error Patterns
"What error patterns are happening in my API collection this week?"
Groups 4xx/5xx errors by status code and URL pattern so you can spot systemic issues.
Search Across Responses
"Search for 'insufficient funds' across all my API responses"
AI Prompts — Just Ask
You don't need to remember prompt names. Just describe what you want in plain English. Your AI assistant knows which RESTK prompt to use.
Scan Your Codebase & Build a Collection
"Scan this codebase and create a RESTK collection with all the API endpoints"
"Generate a collection from this Express project with proper auth and test scripts"
The AI will:
- Scan your routes, controllers, and schemas
- Create an environment with the right variables
- Create a collection with folders for each module
- Create every endpoint with proper URL templates, auth, sample bodies, and test scripts
- Wire up the variable flow — login captures token, create captures ID, downstream requests use them
Supported frameworks: Express, NestJS, Django, Flask, Rails, Spring Boot, Laravel, Go — or auto-detect.
Generate Test Cases
"Write comprehensive tests for the Create User request"
"Generate security-focused test cases for the Payment endpoint"
The AI analyzes the request structure, parameters, and past responses, then suggests test scenarios covering:
- Functional — happy path validation
- Edge cases — empty values, boundary conditions, special characters
- Security — injection patterns, auth bypass attempts
- Error handling — invalid inputs, missing fields
Debug a Failing Request
"Help me debug why the Payment Webhook is failing"
"Why is the Create Order request returning 500?"
The AI gathers the request configuration, recent error responses, headers, and timing — then walks you through likely causes and fixes.
Security Audit
"Do a security review of my Payment Gateway API"
"Check my API collection for missing authentication"
Reviews every endpoint for:
- Missing authentication on sensitive endpoints
- Inconsistent auth patterns across folders
- Missing security headers
- Sensitive data exposed in URLs or query parameters
- HTTP endpoints that should be HTTPS
Each finding is rated by severity with specific fix recommendations.
Optimize a Slow Request
"Help me speed up the List Products endpoint"
"Why is the Search request so slow? Suggest improvements"
Analyzes response history and suggests caching strategies, pagination improvements, compression, field selection, and timeout adjustments.
Audit Agent Readiness
"Is my API collection ready for AI automation?"
"Score my API's quality and completeness"
Evaluates authentication quality, error handling consistency, response format standardization, documentation completeness, and security practices.
Common Workflows
"I just started a new project"
"Scan this codebase and create a RESTK collection"
One sentence — the AI does everything.
"My API is returning errors"
"What errors are happening in my API this week?"
"Help me debug the failing Payment request"
"Show me the error timeline for that request"
"I need to write tests before shipping"
"Write comprehensive tests for the Create Order request"
"Add those tests to the request"
"Run it to verify the tests pass"
"Security review before launch"
"Do a security review of my API"
"Score my API's readiness for production"
"This endpoint is too slow"
"Show me performance stats for List Products over the last month"
"Help me speed up that endpoint"
Seamless Connection
RESTK's connection is designed to be invisible. If you close and reopen RESTK, your AI assistant reconnects automatically — no manual steps needed.
Claude Code and Claude Desktop handle reconnection seamlessly. Cursor users may need to reconnect manually after RESTK restarts.
Approval & Safety
| Action | Approval needed? |
|---|---|
| Browsing collections, requests, history | No |
| Running analysis, generating tests, searching | No |
| Creating or updating requests, folders, collections | Yes — dialog in RESTK |
| Deleting anything | Always yes — even with auto-approve |
| Running cURL commands | Always yes |
Enable auto-approve for non-destructive writes in Settings → AI Integration → Security. Deletes and direct cURL execution always require your explicit approval.
Privacy
- Credentials, API keys, and secrets are never shared with AI
- Only collections you explicitly share are visible
- Response content is sanitized before AI sees it
- All operations are logged in the audit trail
- Everything stays local — no data leaves your machine
For the full privacy architecture, threat model, and FAQ, see MCP Data Privacy & Integrity.
Quick Reference
| Capability | Count |
|---|---|
| Tools | 47 |
| Resources | 10 |
| Prompts | 6 |
| Total capabilities | 63 |
For the full capability reference, see MCP Setup & Usage.