Frequently Asked Questions (FAQ)
General
What is Tydli?
Tydli is a platform that converts OpenAPI specifications into MCP (Model Context Protocol) servers that AI agents can consume. It enables AI assistants like Claude to interact with your APIs seamlessly.What is MCP?
The Model Context Protocol (MCP) is an open standard that enables AI systems to communicate with external data sources and APIs. It provides a standardized way for AI applications to discover and interact with tools and services.Do I need coding experience to use Tydli?
No! Tydli is designed as a no-code platform. If you have an OpenAPI specification for your API, you can generate an MCP server without writing any code.What format should my API documentation be in?
Tydli accepts any format! You have two options:- Any Format (AI-Powered): Upload PDFs, HTML docs, Markdown files, plain text, or paste documentation. Our AI automatically analyzes and generates an OpenAPI specification.
- OpenAPI Direct: If you already have an OpenAPI 3.0+ specification in JSON or YAML format, upload it directly for instant processing.
Account & Authentication
Why do I need to verify my email?
Email verification is required for account security. It ensures that only legitimate users can create and deploy MCP servers on the platform.I didn’t receive the verification email. What should I do?
- Check your spam/junk folder
- Wait a few minutes - emails can take time to arrive
- Try signing up again with the same email to resend the verification
- Contact support if the issue persists
Can I change my email address?
Currently, email changes are not supported through the UI. Contact support for assistance with account changes.What’s the difference between OAuth and JWT authentication?
- OAuth 2.1 with PKCE (Recommended): Best for AI agents like Claude Desktop. Provides secure user consent flow, automatic token refresh, and standardized MCP client integration.
- JWT (Session-based): Direct authentication using your Supabase session token. Good for testing, custom integrations, and API development. Tokens are tied to your login session.
- None (Public): No authentication required - the MCP server is publicly accessible. Usage is still tracked to your account for rate limiting.
Deployments
How long does it take to generate an MCP server?
Typically 20-40 seconds from upload to live deployment, including:- Validation: 2-5 seconds
- Code generation: 10-20 seconds
- Deployment: 5-15 seconds
Why is my deployment stuck in “Generating” status?
Check the deployment logs for error details. Common causes:- Invalid OpenAPI specification
- Network connectivity issues
- Rate limit exceeded
- Large specification taking longer than usual
Can I update an existing deployment?
Currently, deployments are immutable after creation. To make changes:- Upload your updated OpenAPI specification
- Generate a new deployment
- Update your AI agent configuration with the new URL
- Delete the old deployment if no longer needed
How many deployments can I create?
Limits depend on your plan:- Free tier: See your dashboard for current limits
- Paid plans: Higher limits and custom quotas available
Can I make my MCP server public?
Yes! When generating a deployment, select “None” as the authentication method. The server will be publicly accessible, with usage tracked to your account.API Specifications
My OpenAPI spec has authentication. How do I configure it?
After uploading your spec:- Tydli automatically detects authentication requirements
- Configure credentials in the authentication section
- Credentials are encrypted with AES-256 before storage
- The MCP server automatically includes credentials in API calls
Does Tydli support Swagger 2.0?
No, only OpenAPI 3.0+ is supported. You can convert Swagger 2.0 to OpenAPI 3.0 using tools like swagger2openapi.What if my API doesn’t have an OpenAPI spec?
No problem! Use our AI Document Processing feature:- Upload any documentation — PDFs, HTML pages, Markdown files, plain text
- Paste text directly — Copy from websites, emails, or chat messages
- Link to a URL — GitHub READMEs, documentation sites
Can I upload multiple API specs?
Yes! You can create separate deployments for each API. Each deployment gets its own unique URL and authentication configuration.Claude Desktop Integration
Where do I find the Claude Desktop configuration file?
- macOS:
~/Library/Application Support/Claude/claude_desktop_config.json - Windows:
%APPDATA%\Claude\claude_desktop_config.json
{}
Do I need to restart Claude after updating the configuration?
Yes, always restart Claude Desktop after modifyingclaude_desktop_config.json for changes to take effect.
Claude isn’t detecting my MCP server. What should I do?
- Verify the configuration syntax is valid JSON
- Check that all URLs are correct (copy from dashboard)
- Ensure your deployment status is “Ready”
- Restart Claude Desktop completely
- Check Claude Desktop logs for error messages
How do I remove an MCP server from Claude?
- Open
claude_desktop_config.json - Delete the server entry from the
mcpServersobject - Save the file
- Restart Claude Desktop
Rate Limits & Usage
What are the rate limits?
Rate limits vary by plan. Check your dashboard for current limits:- Requests per hour
- Requests per month
- Deployments limit
- AI processing usage
What happens if I exceed my rate limit?
You’ll receive a429 Too Many Requests error. The response includes:
- Current usage
- Limit details
- Time until reset (
Retry-Afterheader)
How is usage tracked?
All MCP server requests are logged and counted toward your usage quota, including:- Successful requests
- Failed requests (still consume quota)
- Requests from any authenticated client
Can I see my usage history?
Yes! View detailed usage analytics in your dashboard, including:- Requests per deployment
- Response times
- Error rates
- Tool invocation logs
Security & Privacy
🔒 For comprehensive security details, see our Security Documentation covering encryption architecture, authentication flows, RLS policies, rate limiting, audit trails, and compliance standards.
How are my API credentials stored?
Credentials are encrypted using AES-256 encryption before storage. They are only decrypted server-side when making API calls - never sent to the client.Who can access my MCP servers?
Access depends on authentication method:- OAuth: Only clients you explicitly authorize
- JWT: Anyone with your JWT token (keep it secret!)
- None: Publicly accessible to anyone
Can I revoke access to an AI agent?
For OAuth authentication:- Go to your deployment settings
- View active OAuth tokens
- Revoke specific tokens
What data does Tydli collect?
Tydli collects:- Account information (email, profile)
- API specifications (stored encrypted)
- Usage analytics (requests, errors, response times)
- Security events (login attempts, rate limits)
- Actual API responses from your endpoints
- Personal data from your APIs
- API credentials in plaintext
Troubleshooting
My MCP server returns 401 Unauthorized
Common causes:- OAuth: Token expired or not yet authorized
- JWT: Invalid or expired session token
- API Credentials: Incorrect credentials configured
- Deployment stopped: Restart the deployment
API calls are failing with “Tool execution failed”
This typically means the target API returned an error. Check:- API credentials are correct
- API endpoint URLs are valid
- Required parameters are provided
- API is reachable and operational
- Deployment logs for the actual API response
How do I debug MCP server issues?
- View Logs: Click “View Logs” in deployment card
- Health Check: Click “Health Check” to test connectivity
- Test with cURL: Copy the cURL command and test directly
- Check API Status: Verify your target API is operational
- Review OpenAPI Spec: Ensure endpoints are correctly defined
Billing & Plans
Is there a free tier?
Yes! The free tier includes limited deployments and monthly requests. Check the dashboard for your current limits.How do I upgrade my plan?
Visit the Billing page in your dashboard to view available plans and upgrade options.Can I downgrade my plan?
Yes, but existing deployments and usage beyond the new plan’s limits will be restricted. Contact support for assistance.Do unused requests roll over?
No, request quotas reset monthly and do not accumulate.Support
How do I get help?
- Documentation: Check User Guide, Troubleshooting
- Dashboard Logs: View deployment logs for error details
- Contact Support: Use the contact form in the application
- GitHub Issues: Report bugs on the project repository
How do I report a bug?
Include the following information:- Description of the issue
- Steps to reproduce
- Expected vs actual behavior
- Deployment ID or request ID (from logs)
- Screenshots if applicable
Can I request a feature?
Yes! Submit feature requests through the contact form or GitHub repository. Include:- Use case description
- Why it’s important to you
- Expected behavior
What are MCP Resources and Prompts?
Resources give AI agents background knowledge about your API — like giving an assistant your employee handbook before their first day. They’re read-only reference materials (API docs, configuration schemas, error code lists) that AI agents can read viaresources/read to make smarter decisions.
Prompts are ready-to-use commands AI agents can execute immediately — like pre-written scripts for common tasks. They’re templates with placeholders (e.g., {{customer_id}}) that get filled in at runtime.
Both are optional but dramatically improve AI accuracy and efficiency with your API.
When should I use Resources vs Prompts?
| Use Resources When… | Use Prompts When… |
|---|---|
| You have static reference data | You need dynamic templates |
| AI needs to read information | AI needs interaction patterns |
| Examples: docs, configs, schemas | Examples: email templates, report formats |
- “Does AI need to read this?” → Resource
- “Does AI need to use this as a template?” → Prompt
How many Resources and Prompts should I create?
Start small — 3-5 of each is usually enough. Focus on:- Resources: The information AI agents ask for most often
- Prompts: Your most common interaction patterns
Can I update Resources and Prompts after deployment?
Yes! Changes take effect immediately. The AI agent will see updated content on the next request. No redeployment needed.Do Resources and Prompts work with all AI agents?
Yes — any AI agent that supports the MCP protocol (Claude, ChatGPT with MCP support, custom agents) can use them. The AI agent automatically discovers available resources and prompts viaresources/list and prompts/list.
What’s the difference between a Resource and just putting info in my API docs?
Resources are AI-readable — they’re structured data the AI can programmatically access and reference during conversations. API docs are human-readable but harder for AI to parse reliably. Resources give AI agents instant, reliable access to exactly the information they need.Can I use Resources for dynamic data?
No — Resources are for static reference data. For dynamic data (like current inventory levels or user profiles), use your API endpoints (Tools). Resources are best for documentation, schemas, rules, and configuration that don’t change frequently.How do I know if my Prompt template is working?
Test it using the API Reference examples. Callprompts/get with sample arguments and verify the rendered output looks correct. You can also check deployment logs to see when prompts are being accessed.