Connect an agent to your data
Everything you make searchable is, from the same account, operable by an AI agent — over MCP, with no glue code. This step is the whole connection: one command.
What MCP does here
MCP (Model Context Protocol) is the standard way an agent plugs into an outside tool. Search Stack runs an MCP server, so once you hand an agent a key, it gains Search Stack's tools — search, suggest, and (with a write-capable key) create, import, edit and delete. The agent decides which tool to call from what you ask it in plain English; you write none of that.
Make a key for the agent
In the console Developers tab, create an API key. What you can do through the agent is exactly what the key allows, so scope it deliberately:
- Read-only (
search-result:read) if the agent should only search and read — safe to hand to an assistant that answers questions. - Write-capable (add
list:create,search-result:create,search-result:edit,search-result:remove) if you want the agent to load and manage data, as this lesson does.
Connect Claude
One command points Claude at your account's MCP server with that key:
claude mcp add --transport http searchstack https://mcp.searchstack.dev/ \
--header "X-API-Key: YOUR_API_KEY"
That's it — no SDK, no server to run. Any MCP-speaking agent works the same way; it's a single server entry in the agent's configuration:
{
"mcpServers": {
"searchstack": {
"url": "https://mcp.searchstack.dev/",
"headers": { "X-API-Key": "YOUR_API_KEY" }
}
}
}
Go deeper: API keys — scopes, permissions, and read-only keys in the reference.