Factagora MCP Server
A Model Context Protocol (MCP) server that connects your AI assistants to your personal knowledge graph. Build your knowledge base once, access it everywhere — Claude, Cursor, and any MCP-compatible client.
Why Factagora?
Stop explaining context to AI every time.
- Switch between projects? No need to re-explain everything.
- Team knowledge? Shared collections everyone can access.
- Fact-check AI responses? Verify against your trusted sources.
Getting Started
2. Installation
Using npx (Recommended)
npx @factagora/mcp-serverGlobal Installation
npm install -g @factagora/mcp-server3. 2. Configure MCP Client
Add to your MCP configuration (.mcp.json or Claude Desktop settings):
Claude Desktop / Claude Code
{
"mcpServers": {
"factagora": {
"command": "npx",
"args": ["-y", "@factagora/mcp-server@latest"],
"env": {
"FACTAGORA_ACCESS_TOKEN": "your_access_token_here"
}
}
}
}With Default Collection (Recommended for single-project use)
Set a default collection to avoid specifying it in every request:
{
"mcpServers": {
"factagora": {
"command": "npx",
"args": [
"-y",
"@factagora/mcp-server@latest",
"--collection-id=your-collection-uuid"
],
"env": {
"FACTAGORA_ACCESS_TOKEN": "your_access_token_here"
}
}
}
}Usage
1. Get Your Collections
After connecting, first get the list of collections you have access to.
get_collections— List all collections you have access toget_collection_detail— Get detailed information about a collection
2. Search or Verify
Two actions are available on your selected collection.
search_collection— Search for factblocks using natural language queriesdeepverify— Verify a claim against your knowledge base
Configuration Options
Environment Variables
| Variable | Required | Description |
|---|---|---|
FACTAGORA_ACCESS_TOKEN | Yes | Your Factagora API access token |
Command Line Arguments
| Argument | Description |
|---|---|
--collection-id | Set default collection for all operations |
