The agent acts in your backend.
Cancel a subscription, claim a listing, unlock an account: the actions that matter live in your system, not ours. Animam calls your endpoint with the visitor’s verified identity — and never runs your code.
Three ways to plug your system in
All declarative: you provide a URL and a contract, never a program to run.
Verified action
A signed POST to YOUR endpoint, fired by the agent when the conversation calls for it. You receive the visitor’s verified email and the parameters you declared — nothing else.
Your own MCP server
Plug in an MCP server you already run: its tools show up in the agent, discovered on the fly and cached for 5 minutes.
Webhook on a form
The simplest one: every submission is posted to a URL of your choice, on top of the email. Useful when your system already knows what to do with a lead.
What the platform guarantees
An agent that can act on an account is an attack surface. Here is what is enforced on every call, with nothing to configure on your side.
Identity is never dictated by the model
The email sent to your backend is the one the server verified (one-time code or JWT). If the model proposes a different one, it is dropped.
Only your parameters get through
The model can only fill the parameters declared in the config. It cannot invent one, nor reach another field of your API.
No calls into a private network
The URL is validated before every call, DNS resolution included: loopback, private ranges and cloud metadata endpoints are refused — including if DNS changes between the check and the call.
Your secret is encrypted at rest
It is encrypted with AES-256-GCM on write, decrypted only at call time, and never readable back through an API.
The response is filtered before the model
Secret-looking keys are stripped from your response, strings are capped, and the call is abandoned after 5 seconds.
Your code stays yours
Nothing you write runs on our infrastructure. You hand us a URL and a contract, not a program. Your server decides.
One request to declare it
A verified action is configured through the API, with a key carrying tool write access. Two agents not to confuse here: your personal agent (Claude, Hermes, ClawBot) configures Animam over MCP — this is the one tool type it may NOT create, because it carries a secret, and a secret is not delegated to a model. Your site agent is the one that uses it: it fires the call during the conversation. There is no dedicated dashboard screen either; we would rather say so than let you find out.
# One POST creates the tool. The secret is encrypted at rest (AES-256-GCM).
curl -X POST https://api.animam.ai/tenants/acme/tools \
-H "X-API-Key: $ANIMAM_API_KEY" \
-H "Content-Type: application/json" \
-d '{
"name": "claim_listing",
"description": "Claim ownership of a listing for the verified visitor.",
"type": "VERIFIED_ACTION",
"config": {
"url": "https://your-backend.example/api/claim",
"secret": "whatever-your-backend-checks",
"params": [{ "name": "listing_id", "required": true }],
"messages": {
"ok": "Done — check your inbox for the confirmation.",
"already_claimed": "That listing is already claimed."
},
"failureMessage": "I could not reach the system. Try again shortly."
}
}'The visitor must have proven their email before any action fires: the agent walks them through a one-time code inside the conversation. An action can therefore never be triggered by someone claiming to be someone else.
Included from Builder up
Verified actions, external MCP servers and custom webhooks are open on the Builder, Pro and Agency plans.