MCP Integration
Connect your AI assistant to Feedback1 using the Model Context Protocol (MCP).
What you can do
MCP lets AI assistants pull live context from your Feedback1 workspace:
- Search feedback by keyword or semantic meaning
- Find and list roadmap features
- Look up CRM accounts and contacts
- Get feedback statistics and top-voted features
- Retrieve specific items by ID
Connect with OAuth (recommended)
Cursor and Claude can connect with just the server URL — no API key in config. Your client discovers OAuth automatically, opens a browser to sign in to Feedback1, and you pick which product to authorize.
- In Cursor or Claude Desktop MCP settings, add a remote server with URL
https://app.feedback1.ai/api/mcp/sse - Complete sign-in when prompted (Feedback1 account + product selection)
- Start using MCP tools in your assistant
Cursor / Claude config (URL only)
{
"mcpServers": {
"feedback1": {
"url": "https://app.feedback1.ai/api/mcp/sse"
}
}
}In-app setup: Settings → MCP
Alternative: API key (manual Bearer)
Use an API key if your client does not support OAuth discovery, or for mcp-remote / Claude Desktop command-based configs.
Requirements (API key path)
- A Feedback1 workspace API key — create one in Settings → API Keys
- Node.js 18+ for Claude Desktop (uses the
mcp-remoteproxy vianpx) - SSE endpoint:
https://app.feedback1.ai/api/mcp/sse
1. Get your API key
- Log in to app.feedback1.ai
- Go to Settings → API Keys and create a key
- Copy the key (format:
apikey_xxxorf1_live_xxx)
Logged-in users can also use the in-app MCP settings page to copy a ready-made config.
2. Connect Claude Desktop
Edit your Claude Desktop config file:
- macOS:
~/Library/Application Support/Claude/claude_desktop_config.json - Windows:
%APPDATA%\Claude\claude_desktop_config.json
Basic configuration (all products)
{
"mcpServers": {
"feedback1": {
"command": "npx",
"args": [
"mcp-remote@latest",
"https://app.feedback1.ai/api/mcp/sse",
"--header",
"Authorization: Bearer YOUR_API_KEY"
]
}
}
}Replace YOUR_API_KEY with your actual key, then restart Claude Desktop.
3. Product-scoped configuration (recommended)
If you have multiple products, set a default product so queries are automatically scoped:
{
"mcpServers": {
"feedback1": {
"command": "npx",
"args": [
"mcp-remote@latest",
"https://app.feedback1.ai/api/mcp/sse",
"--header",
"Authorization: Bearer YOUR_API_KEY",
"--header",
"X-Feedback1-Product-ID: YOUR_PRODUCT_ID"
]
}
}
}Replace YOUR_PRODUCT_ID with your product's public ID (e.g. ABC123). Find it under Settings → Products.
4. Connect Cursor
In Cursor, open MCP settings and add a server. Use the same SSE URL and Bearer auth. Example config:
{
"mcpServers": {
"feedback1": {
"url": "https://app.feedback1.ai/api/mcp/sse",
"headers": {
"Authorization": "Bearer YOUR_API_KEY"
}
}
}
}Alternatively, use the same npx mcp-remote args as Claude Desktop if your Cursor version expects a command-based server.
Available tools
Feedback1 MCP exposes search, list, retrieval, analytics, and CRM tools:
- Discovery:
list_products - Search:
search_feedback,search_features,search_accounts,global_search - Retrieval:
get_feedback,get_feature - Lists:
list_feedback,list_features,list_accounts,list_contacts - Analytics:
get_feedback_stats,get_top_voted_features - CRM:
get_account,get_contact, and related feedback/feature lookups
Reference resources include feedback types, feedback states, feature states, and feature priorities (feedback1://… URIs).
Example prompts
- “Search for feedback about export functionality”
- “What features are currently in progress?”
- “Show feedback statistics for the last 30 days”
- “Find enterprise customers in healthcare”
Troubleshooting
Invalid API key
Verify the key is active in Settings → API Keys and that you copied the full value.
ReadableStream is not defined
You need Node.js 18+. Update Node or point command to a Node 18+ binary.
No results / product_id is null
Add the X-Feedback1-Product-ID header, or run list_products first and pass product_id in queries.
Connection timeout
Confirm the URL is https://app.feedback1.ai/api/mcp/sse and restart your AI client after config changes.
Security & rate limits
- OAuth tokens and API keys are tenant-scoped — they only access your organization's data
- OAuth access is bound to the product you approve during consent
- All traffic is encrypted over HTTPS
- 100 requests per minute per API key; search results paginate (max 50 per query)
REST API
For direct HTTP integrations (Zapier, custom scripts, webhooks), see the Public API docs.
Support
Questions? Email [email protected].