Before each send, your AI agent submits the email to the API (one call, or 2 lines with the SDK).
Qorami.AI
The guardrail for AI agents that take action.
Before an AI agent sends a sensitive email or commits your business, Qorami verifies, decides and secures the action.
Not an email-sending tool. The layer that decides whether an agent may send, must ask for confirmation, or should be blocked — for sales, support and ops teams whose AI agents send email.
AuthorizeConfirmBlockSecure
DecisionA confirmer
- Recipient verified
- Risk policy evaluated
- Audit trail created
- Audit trail per workspace
AI agents are not short on intelligence. They are short on guardrails at the moment of acting.
How it works
An LLM judge reads intent in any language + deterministic rules. Every decision is traced.
Send, ask a human, or block. The agent obeys; you stay in control.
SendAsk a humanBlock
Agent API
An API that AI agents can call before they act.
The public demo analyzes without an account. The agent endpoint is API-key protected: it verifies the email, returns an executable instruction and creates a persistent trace.
send— the agent sends the email.request_human_confirmation— the agent pauses and asks a human to approve.do_not_send— the agent does not send.
Agent-ready verification
POST https://qorami.fr/api/verify-email
x-qorami-api-key: <votre_cle>
content-type: application/json
{
"agentId": "outbound-agent",
"recipient": "client@example.com",
"subject": "Offre commerciale",
"body": "Bonjour...",
"policyProfile": "sales"
}
Response:
{
"verification": { "decision": "A confirmer", "actionId": "act_xxxx" },
"nextAction": { "type": "request_human_confirmation" }
}Agent side — ~10 linesconst r = await fetch('/api/verify-email', {
method: 'POST',
headers: { 'x-qorami-api-key': KEY, 'content-type': 'application/json' },
body: JSON.stringify(email),
}).then((res) => res.json())
switch (r.nextAction.type) {
case 'send': mailer.send(email); break
case 'request_human_confirmation': queueForHuman(email, r.verification); break
case 'do_not_send': drop(email, r.verification.reasons); break
}Production-ready
Built for production, not a demo.
Workspace isolation
Each team only sees and acts on its own workspace. API keys are revocable.
Automatic retention
Email bodies redacted after 7 days, actions deleted after 30. Audit metadata stays.
Audit + traceability
Every response carries a request-id; every action is traced and queued for human approval.
Limits & idempotency
Per-key rate limiting and idempotency: no abuse, no double action if an agent retries.
Security & Integration
Secure your AI agents in 3 minutes
Whether you use our SDK (JS/Python), an agent tool (LangChain, LlamaIndex), or an MCP server (Claude Desktop, Cursor), Qorami steps in before sending to protect your communications without slowing down your workflows.
Get an API key by creating a free account.
Add the SDK or the MCP server to your agent.
Define your filtering policy in the console.
What Qorami protects for you:
- ✓ Your sensitive data and API keys against accidental leaks
- ✓ Your finances against unapproved commercial commitments
- ✓ Your reputation against sending errors and prompt injections
A control layer for real actions.
Qorami sits between the agent and your business tools. Each action gets a clear decision: allow, ask for confirmation, or block.
Validate
Check the recipient, content, context and company rules.
Trace
Keep a clear history of what the agent tried to do and why.
Pay-as-you-go
Only pay for the emails you analyze, with no fixed monthly subscription. Blocking risky emails remains free.
Built for your agents’ business emails.
It’s concrete, frequent and risky: a bad agent email can cause an information leak, an unvalidated commercial promise or a wrong recipient.
Your agent prepares a send
Qorami analyzes the risk
Your policy allows, suspends or blocks
You keep a complete audit trail
A simple model: pay as you go.
Our pricing is transparent and direct: no complex subscriptions, you only pay for the email actions you secure.
1 protected action = 1 credit. 100 free verifications to get started.
verifications included to test Qorami on a real AI-agent workflow.
Create account500 protected actions, about 0.02 EUR per traced verification.
Buy3000 verifications in a single payment — about 0.016 EUR per verification.
BuyBusiness policies
The same email does not carry the same risk depending on context.
Each agent picks a profile. Thresholds and signals change by business context, making it more useful than a generic filter.
Standard business emails, cautious without being too strict.
Prices, discounts, commercial promises and signing urgency.
Unhappy customers, refunds, escalations and relational tone.
Contracts, payments, confidentiality and financial commitments.
Frequently asked questions
Does Qorami send the emails?
No. Qorami decides (send / ask a human / block); your agent does the sending. We are the control point before the send, never the sender.
How is it better than a keyword filter?
An LLM judge reads intent in any language (even with no obvious keyword), with a deterministic fallback and real-secret detection (API keys, tokens). Every decision is traced.
How do I integrate Qorami?
One API call before sending, or 2 lines with the SDK (JS/Python) or the LangChain tool. The agent obeys nextAction: send / request_human_confirmation / do_not_send.
What does it cost?
1 traced verification = 1 credit (a block is free). Free credits to test, then prepaid packs. No card required to start.
What about email privacy?
Email content is redacted and auto-deleted per a retention policy; only decision metadata is kept for audit.
Connect Qorami to a real email agent.
Create a free workspace, copy the paste-ready command, then send your first email through /api/verify-email. No card required.
Not ready to create an account? Tell us your use case and we’ll get back to you.