Install LoginLlama in One Prompt
Add LoginLlama to your app with one prompt in Claude Code, Cursor, or Codex, plus an MCP server so your agent can answer questions about your logins.
Loginllama just got a lot easier to start with.
Until now, adding suspicious login detection to your app followed the same tired script every API asks of you: install a package, copy a snippet, find where to paste your API key, wire it into your auth flow, and only then get to test it. None of it is hard. All of it is friction. And friction is the enemy of "let me just try this for five minutes."
So we rebuilt the first five minutes around the tool most developers already have open: their coding agent.
Install with one prompt
If you use Claude Code, Cursor, or Codex, you no longer integrate Loginllama by hand. You hand it the prompt and let it do the work:
- It detects your language and framework and installs the right SDK (Node, Python, or PHP).
- It reads our live docs, wires the risk check in after your auth step, and branches on the result — block, step up to MFA, or allow.
- It makes sure the key lives in an environment variable, never in your code.
- It fails open by design, so a hiccup on our side never locks out your users.
Grab the prompt from the homepage or the docs, paste it into your agent, and you're scoring logins before your coffee's cold. Prefer to do it by hand? The SDKs haven't gone anywhere — they're one import away.
Ask your logins anything
The other half of an agent-first product is letting the agent read your data, not just write code. So Loginllama now ships an MCP server.
Connect it to your agent with your API key and you can ask, in plain English:
"What's my average login risk score for the Chrome browser?"
"Which country had the riskiest logins this week?"
"Show me the ten most recent high-risk logins."
Behind the scenes those questions hit a small set of read-only tools, each scoped to your team's data. No dashboards to dig through, no CSVs to export — you ask, it answers. Setup lives in the docs.
Let an agent get its own key
Pasting an API key works, but it's still a manual step. So we added a proper device-authorization flow (the same kind of grant your TV uses to log into a streaming app). An agent can request access, you get a short code, you approve it once on a screen that shows exactly what you're granting and for which team — and the agent receives a scoped key. No copy-pasting secrets around. You can revoke that key any time from your API settings.
Docs that machines can read
We also published an llms.txt: a clean,
machine-readable summary of the API, the SDKs, the install prompt, and the MCP
server. It's the file your agent reads so it gets the integration right the first
time, without scraping a marketing page.
Why we're doing this
Login security shouldn't be a project. It should be a decision you make in an afternoon and then forget about, because it just works. Making the install a single prompt — and making your analytics a single question — is how we get there. It also means you can genuinely try Loginllama with almost no commitment, on free checks, which is exactly how we think developer tools should be evaluated.
This is the first wave of agent-first work, and there's more coming: richer MCP tools, per-framework recipes, and tighter team controls around what agents can do on your behalf.
If you've been meaning to add login protection but never found the hour, this is your sign. Start with 100 free checks, no card — and let your agent do the wiring.