Connect an MCP client
Use the hosted server for account access. Local stdio setup is for developers with a repository checkout.
Claude remote connectors
Claude supports remote custom connectors separately from local Desktop servers. Add the hosted URL and configure the request header required by your Fenceline key:
https://mcp.fenceline.ai/mcp
X-API-Key: YOUR_API_KEY Follow Claude’s remote connector instructions for your plan and organization. Request-header support and organization policies determine the setup available to you. This guidance was checked on October 25, 2026.
Other remote clients
Use your client’s documented format for the URL and headers. This example illustrates the connection values; configuration keys vary by client.
{
"mcpServers": {
"fenceline": {
"url": "https://mcp.fenceline.ai/mcp",
"headers": { "X-API-Key": "YOUR_API_KEY" }
}
}
} Discovery of server resources works without a key. Tool discovery and calls require one. Keep keys out of checked-in files.
Desktop compatibility bridge
For a local stdio configuration, the dashboard also provides an mcp-remote bridge to
the authenticated SSE endpoint. This option requires Node.js and permission to run the bridge. Use
the configuration supported by your installed client.
{
"mcpServers": {
"fenceline": {
"command": "npx",
"args": [
"-y",
"mcp-remote",
"https://mcp.fenceline.ai/mcp/sse",
"--header",
"X-API-Key:YOUR_API_KEY"
]
}
}
} See Claude’s local server documentation for installation and organization controls.
Local server development
These steps require a repository checkout and the server’s environment dependencies. Review servers/mcp/env.example and supply credentials through your approved local secret mechanism.
bash scripts/npm-pinned.sh --prefix servers/mcp ci --ignore-scripts --prefer-offline
npm --prefix servers/mcp run build {
"mcpServers": {
"fenceline-local": {
"command": "node",
"args": ["/absolute/path/to/fenceline/servers/mcp/dist/index.cjs"],
"env": {
"MCP_TRANSPORT": "stdio",
"FENCELINE_CONTRACTOR_ID": "YOUR_CONTRACTOR_ID"
}
}
}
} Use an absolute path. Start with discovery and a read operation before testing tools with side effects.