Skip to main content

​
Your AI Will Know Your Entire Codebase in 5 Minutes

Right now: Your AI hallucinates patterns. You copy-paste docs every session. New developers ask the same questions for weeks. 5 minutes from now: Your AI assistant has instant access to every README, wiki page, and architecture doc across all your repositories. This guide gets you there—fast.

​
What You’ll Accomplish

Connect GitHub

2 minutesOAuth connection → automatic repository discovery

Index Documentation

1-3 minutes (automatic)AI processes all your docs in the background

Configure AI Tool

2 minutesAdd ULPI MCP server to Claude Code, Cursor, or 40+ other tools

Ask First Question

Instant answers“How do we deploy?” → Your AI knows immediately
Total time: 5-7 minutes. Most of it is waiting for automatic indexing.

​
Prerequisites

You’ll need:

GitHub Account

Sign up free at github.comWorks with personal or organization accounts

AI Coding Assistant

Any MCP-compatible tool:
  • Claude Desktop (recommended)
  • Cursor
  • Windsurf
  • VSCode with Continue/Cline
  • See all 40+ tools
Repository requirements:
  • Public or private repositories (both supported)
  • GitHub, GitLab, Bitbucket, or Gitea
  • Documentation in Markdown, README files, wikis, or code comments

​
Step 1: Create Your ULPI Account

Stop wasting time searching. Let’s fix this now.
1

Visit app.ulpi.io

Go to app.ulpi.io and click Sign Up
No credit card required for signup. Choose your plan after you see the value.
2

Authenticate with GitHub

Click Continue with GitHub to authenticate via OAuthWhat we request:
  • ✅ Read-only access to repositories
  • ✅ Webhook access (for auto-sync)
  • ❌ We never write to your code
  • ❌ We never modify repositories
Using GitLab, Bitbucket, or Gitea? Select “Other Git Provider” and follow the connection wizard.
3

Choose Your Plan

Select a plan based on your team size:
Perfect for solo developers or small teams:
  • 5 repositories
  • 100,000 tokens/month (~15-20 searches/day)
  • Semantic search across all docs
  • Auto-sync on every push
  • MCP integration (40+ AI tools)
  • Email support
Best for: 1-3 developers
Compare all plans in detail →
Not sure? Start with Starter. Upgrade anytime. We’ll prorate the difference.Bundle discount: Save 15-20% by combining Documentation with Skills, Memory, Coordination, or Hooks.
4

Complete Billing

Enter payment details via Stripe (secure, PCI-compliant)Guarantees:
  • 🔒 Bank-level security (PCI-DSS Level 1)
  • 💳 30-day money-back - cancel anytime, full refund
  • 🔄 Flexible billing - change plans monthly
  • 📊 Transparent usage - see token consumption in real-time
Your account is live immediately after payment confirmation.
Estimated time: 2 minutes

​
Step 2: Connect Your Repositories

Make your documentation searchable by AI.
1

Navigate to Repositories

In the ULPI dashboard:
  1. Click Repositories in the left sidebar
  2. You’ll see the repository connection screen
If you just signed up, you’ll be automatically directed here.
2

Select Repositories to Connect

Choose which repositories to make searchable:
Best for testing or specific projects:
  1. Browse your repositories in the list
  2. Check the boxes next to repositories with documentation
  3. Click Connect Selected (green button)
Recommended for first-time setup: Start with 2-3 repos to see ULPI in action, then add more.
3

Grant Permissions

ULPI requests authorization to:
What we read:
  • Documentation files (.md, .mdx, README.*)
  • Wiki pages (optional, you can disable)
  • Code comments (optional, disabled by default)
What we don’t read:
  • Source code (unless you enable comment indexing)
  • Secrets, .env files
  • Build artifacts, node_modules/
  • Binary files
Why we need this:
  • Automatic re-indexing on git push
  • Keep documentation always up-to-date
  • No manual “sync” button needed
Triggered on:
  • Push to main/master branch
  • Pull request merges
  • Wiki updates
  • Manual trigger from dashboard
Security guarantee:
  • ❌ Cannot modify files
  • ❌ Cannot create commits
  • ❌ Cannot push changes
  • ❌ Cannot delete repositories
Read-only access only. Revoke anytime from GitHub settings.
Click Authorize ULPI to proceed.
4

Wait for Initial Indexing

ULPI automatically discovers and indexes your documentation:
Automatically included:
  • ✅ README.md, README.txt (all directories)
  • ✅ docs/ and documentation/ directories
  • ✅ All .md and .mdx files
  • ✅ CONTRIBUTING.md, CHANGELOG.md
  • ✅ Architecture Decision Records (ADRs)
  • ✅ .github/ documentation
Optionally included (configure in settings):
  • 🔧 Wiki pages (disabled by default, can enable)
  • 🔧 Code comments and docstrings (disabled by default)
  • 🔧 Jupyter notebooks (.ipynb)
  • 🔧 .txt files in doc directories
Automatically excluded:
  • ❌ node_modules/, vendor/, packages/
  • ❌ Build directories (dist/, build/, out/)
  • ❌ .git/ directory
  • ❌ Binary files (images, PDFs, etc.)
  • ❌ Large files (>1MB)
Custom exclusions: Create a .ulpiignore file (works like .gitignore)
Estimated time: 2 minutes to connect, 1-15 minutes for automatic indexing
If you don’t see your repository:
  1. Check GitHub OAuth permissions:
  2. Organization settings:
    • Org admin must approve third-party apps
    • Ask your GitHub org admin to approve ULPI
    • Guide for admins →
  3. Repository visibility:
    • Private repos require explicit OAuth permission
    • Re-authenticate and grant access to private repositories
  4. Still not working?
    • Contact support: support@ulpi.io
    • Include: repository URL, GitHub username, error messages

​
Step 3: Create an API Key

API keys connect your AI assistant to ULPI’s documentation search.
1

Navigate to API Keys

In the ULPI dashboard:
  1. Click Settings in sidebar
  2. Select API Keys tab
  3. You’ll see the API key management screen
2

Click Create API Key

Click the Create API Key button (green button, top right)
3

Configure Key Settings

Fill in API key details:
Name: Documentation MCP Server(For identification—name it after where you’ll use it)Examples:
  • “Claude Desktop - MacBook Pro”
  • “Cursor - Work Laptop”
  • “Team Shared Key”
Environment: live (production) or test (development)
  • live: Use in real AI assistants (recommended)
  • test: Use for testing, development, experiments
Start with live environment unless you’re developing integrations.
4

Copy Your API Key

⚠️ CRITICAL: This is shown ONLY ONCE!
  1. Copy the API key immediately
  2. Store it securely:
    • Password manager (1Password, LastPass, Bitwarden)
    • Environment variable in your shell config
    • Secure note in your IDE
API key format: ulpi_live_sk_1234abcd...
Never commit API keys to version control!Add to .gitignore:
# ULPI API Keys
.env
.env.local
**/mcp_config.json
claude_desktop_config.json
Lost your key? You can’t recover it, but you can:
  1. Revoke the old key
  2. Create a new one
  3. Update your AI tool config with the new key
Estimated time: 1 minute

​
Step 4: Configure Your AI Assistant

Connect ULPI to your favorite AI coding tool. Choose your AI assistant below and follow the setup: Estimated time: 2 minutes
Verify everything works!
1

Open Your AI Assistant

Launch the AI tool you configured in Step 4:
  • Claude Desktop
  • Cursor
  • VSCode with Continue/Cline
  • etc.
2

Ask a Documentation Question

Try these example queries:
"How do I set up the local development environment?"
Ask in plain English! ULPI understands natural language—no need for exact keywords.
3

Verify Results

Your AI assistant should:✅ Search ULPI automatically (you won’t see this, it happens behind the scenes)✅ Return relevant documentation excerpts from your repositories✅ Include source file links like:
Based on docs/setup.md:15-45:
To set up your development environment...
✅ Provide context-aware answers using YOUR project’s patterns✅ Cite multiple sources if the answer spans multiple docs
Success looks like: AI responds with specific details from your actual documentation, not generic advice.
4

Monitor in Dashboard

See searches in real-time:
  1. Go to ULPI dashboard → Activity tab
  2. You’ll see:
    • Recent searches
    • Which repositories were queried
    • Token usage
    • Response times
Example activity log:
10:34 AM - Search: "How do we deploy?"
           → Found in: infrastructure/deployment.md
           → Tokens used: 2,340
           → Response time: 45ms

10:31 AM - Search: "API authentication methods"
           → Found in: backend-api/README.md, docs/auth.md
           → Tokens used: 1,890
           → Response time: 38ms
Estimated time: 1 minute

​
Troubleshooting

Symptom: AI gives generic answers instead of searching your docs.Fixes:
  1. Check MCP server connection:
    • Claude Desktop: Look for 🔌 icon in bottom toolbar
    • Cursor: Check Settings → MCP Servers shows “Connected”
    • VSCode: Continue/Cline should show “ulpi-docs” in status
  2. Verify API key format: 
    {
  "env": {
    "ULPI_API_KEY": "ulpi_live_sk_abc123..."  // ✅ Correct
  }
}
    Common errors: 
    "ULPI_API_KEY": " ulpi_live_sk_abc123"  // ❌ Leading space
"ULPI_API_KEY": "ulpi_test_sk_abc123"   // ❌ Wrong environment
"ULPI_API_KEY": "'ulpi_live_sk_abc123'" // ❌ Extra quotes
  3. Check MCP server logs:
    • macOS: ~/Library/Logs/Claude/mcp.log
    • Windows: %APPDATA%\Claude\Logs\mcp.log
    • Linux: ~/.config/Claude/logs/mcp.log
    Look for errors like: 
    ERROR: Authentication failed - invalid API key
ERROR: ULPI_API_KEY environment variable not set
ERROR: Network error connecting to ULPI API
  4. Restart AI assistant:
    • Fully quit (don’t just close window)
    • Reopen and try again
    • Config changes only load on full restart
Symptom: Repository stuck at “Indexing…” for >30 minutes.Normal indexing times:
  • Less than 100 files: 30 seconds - 2 minutes
  • 100-1,000 files: 2-5 minutes
  • 1,000-10,000 files: 5-15 minutes
  • 10,000+ files: 15-30 minutes
If it’s taking longer:
  1. Check repository size:
    • Dashboard → Repositories → Click repo name
    • See “Files discovered” count
    • Large repos (10,000+ files) can take 30-45 minutes
  2. Check for errors:
    • Look for ⚠️ warning icon next to repository
    • Click for error details
    • Common issues:
      • Binary files causing slowdown (auto-excluded, but takes time to detect)
      • Network issues accessing GitHub
      • Very large markdown files (>5MB)
  3. Force re-index:
    • Dashboard → Repositories → Click ⟳ Re-index button
    • This restarts the indexing process
    • Sometimes resolves stuck states
  4. Contact support:
    • If stuck >1 hour: support@ulpi.io
    • Include: repository name, size, approximate file count
Symptom: AI says “I couldn’t find documentation about that” or returns empty results.Possible causes:
  1. Indexing not complete:
    • Dashboard → Repositories → Check status
    • Wait for ”✅ Indexed” status
    • Usually takes 2-5 minutes
  2. Documentation doesn’t exist:
    • Verify your repos actually have docs about that topic
    • Check if docs use different terminology
    • Example: You ask about “authentication” but docs say “auth” or “login”
  3. Wrong repository scope:
    • API key might be scoped to specific repos
    • Dashboard → Settings → API Keys → Check scopes
    • Expand scope or use different key
  4. Query too vague:
    • ❌ “How does it work?”
    • ✅ “How does authentication work?”
    • ✅ “Explain our Redis caching strategy”
    • Be specific!
  5. Files excluded by default:
    • Check .ulpiignore file in repository
    • Dashboard → Repositories → Click repo → See “Excluded patterns”
    • Adjust exclusions if needed
Symptom: “Invalid API key” or authentication errors.Fixes:
  1. Verify key format: 
    ulpi_live_sk_1234567890abcdef...  // ✅ Correct (live)
ulpi_test_sk_1234567890abcdef...  // ✅ Correct (test)
ulpi_sk_1234567890abcdef...       // ❌ Missing environment
sk_1234567890abcdef...             // ❌ Missing ulpi_ prefix
  2. Check environment match:
    • API key environment (live vs test)
    • MCP server environment
    • Dashboard shows environment per key
  3. Verify key not revoked:
    • Dashboard → Settings → API Keys
    • Check key status is “Active” (not “Revoked”)
    • If revoked, create a new key
  4. Test API key directly: 
    curl -H "Authorization: Bearer ulpi_live_sk_YOUR_KEY" \
     https://api.ulpi.io/v1/search?q=test
    Expected response: 
    {
  "success": true,
  "results": [...]
}
    Error response: 
    {
  "error": "Invalid API key"
}
Symptom: “You’ve exceeded your monthly token limit” error.Solutions:
  1. Check current usage:
    • Dashboard → Usage & Billing
    • See tokens used vs. plan limit
    • View usage by day/week/month
  2. Upgrade plan:
    • Starter (100k) → Professional (500k)
    • Professional → Enterprise (2M)
    • Settings → Billing → Change Plan
  3. Buy token overage:
    • $20 per 100,000 additional tokens
    • No service interruption
    • Automatically billed at end of month
    • Dashboard → Usage → “Buy More Tokens”
  4. Set usage alerts:
    • Dashboard → Settings → Usage Alerts
    • Get notified at 50%, 75%, 90% of limit
    • Prevents surprises
  5. Optimize searches:
    • Be more specific in queries (returns less data)
    • Scope searches to specific repositories
    • Avoid overly broad questions
Symptom: “Access denied” when trying to search private repositories.Fixes:
  1. Re-authenticate with GitHub:
    • Dashboard → Settings → Integrations
    • Click “Reconnect GitHub”
    • Grant access to private repositories
  2. Check organization approval:
    • GitHub org admin must approve ULPI app
    • Ask admin to:
      1. Go to github.com/organizations/YOUR_ORG/settings/oauth_application_policy
      2. Find “ULPI Documentation”
      3. Click “Grant access”
  3. Verify API key scope:
    • Dashboard → Settings → API Keys
    • Click key name → View scopes
    • Ensure key has access to that repository
  4. Repository permissions:
    • You must have at least “Read” access to repo in GitHub
    • Org admins can grant access via GitHub teams
Symptom: Connection timeouts or “Cannot reach ULPI API” errors.Enterprise network fixes:
  1. Whitelist ULPI domains: 
    api.ulpi.io
app.ulpi.io
*.ulpi.io
  2. Allow outbound HTTPS (port 443):
    • ULPI MCP server needs to reach api.ulpi.io:443
    • Check corporate firewall rules
  3. Proxy configuration: 
    {
  "env": {
    "ULPI_API_KEY": "ulpi_live_sk_...",
    "HTTPS_PROXY": "http://proxy.company.com:8080",
    "HTTP_PROXY": "http://proxy.company.com:8080"
  }
}
  4. VPN conflicts:
    • Try disconnecting VPN temporarily
    • If works without VPN, configure VPN to allow ULPI
  5. Test connectivity: 
    curl -I https://api.ulpi.io/health
    Expected: HTTP 200 OK Error: Connection timeout or refused
Still stuck? Email support@ulpi.io with:
  • Error message screenshot
  • MCP server logs
  • API key (first/last 4 characters only)
  • AI tool and OS you’re using

​
What’s Next?

Advanced Search Features

Master semantic search, filters, and query optimizationLearn: Cross-repo search, branch filtering, date ranges, file type filters

How It Works (Technical Deep Dive)

Understand the architecture behind ULPI’s semantic searchLearn: Vector embeddings, Typesense, chunking strategies, indexing pipeline

Repository Management

Add more repositories, configure indexing, manage webhooksLearn: Multi-repo strategies, .ulpiignore, custom exclusions, wiki integration

API Integration

Integrate ULPI directly into your applications via REST APILearn: Direct API access, custom search UIs, workflow automation

​
Success Story

Your team’s documentation is searchable now. Time to see the same transformation.
Need help? We’re here for you:Average response time: Under 2 hours during business hours