CRM security at agentlyleads
What actually protects your data, stated plainly. No claim on this page is a certification we have not earned; where a buyer needs paperwork instead of a description, we say so and link to it.
Every workspace is isolated at the data layer
Your workspace id comes from your session, never from anything a client sends, and it is included in every database query the application makes, not checked once at the edge and trusted after that. That same rule is what keeps the MCP connector from letting an assistant see or change data outside the workspace it authenticated into.
Passwords and sessions
Passwords are hashed with bcrypt at cost 12 before they ever touch the database; the plaintext you type is never stored. Signing in issues a database-backed session token in an HTTP-only cookie with a 30-day lifetime, so a session can be revoked server-side rather than relying on the cookie alone to expire.
What the AI is and is not allowed to do
The connector authenticates with OAuth 2.1 and PKCE, and you approve access on a consent screen before it can see anything. From there, reads are open: an assistant can search and summarize freely. Writes are confirmation-gated, outbound email is always saved as a draft for a person to approve before it sends, and there are no delete tools at all. Read more on the connector page.
Credentials at rest
Third-party credentials we store on your behalf, mailbox connections and AI provider keys, are encrypted at rest with AES-256-GCM rather than kept as plain text in the database.
The application itself
Security headers and a Content Security Policy are set at the framework level on every response, not added page by page where one could be missed.
Infrastructure
The application runs on AWS inside a private VPC. The database is not reachable from the public internet; even our own deployment migrations run as a one-off job inside the VPC because that is the only place that can reach it.
What we do not claim
No SOC 2, no ISO 27001, no HIPAA attestation, no penetration test, no bug bounty. If your process needs one of those, ask, but do not take our word for something we have not had verified. What we do have is a data processing agreement and a subprocessor list, both public, and both a more honest starting point than a badge on a marketing page.
Common questions
- Is agentlyleads SOC 2 or ISO 27001 certified?
- No, and we would rather say that plainly than let the question go unanswered. There is a data processing agreement and a subprocessor list, both public, which cover what most buyers actually need before signing.
- Has agentlyleads had a penetration test or run a bug bounty?
- No. We are not going to claim testing that has not happened. The measures on this page, workspace isolation, password hashing, encrypted credentials, and a private database, are what is actually in place today.
- Can the AI assistant see or change another workspace's data?
- No. The workspace id an AI action runs under comes from the authenticated session, never from anything the assistant or a client sends, and it is applied on every database query rather than checked once at the edge.