Hosted in Singapore
The database runs on Neon in AWS Singapore and the application on DigitalOcean in Singapore. Email goes out through Amazon SES in the same region.
Security
Agents act with your credentials, so three things must never leak: your data, your credentials and your authority. Each has its own controls, and none of them relies on an AI model behaving well.
Your data
The database runs on Neon in AWS Singapore and the application on DigitalOcean in Singapore. Email goes out through Amazon SES in the same region.
Row-level security in the database binds every request to one account, so a query can't return another customer's rows even if our code forgets a filter. Automated tests try to read and write across accounts to prove that isolation holds.
Every change and every agent action is written to an append-only log. Each entry is chained to the one before it, so any tampering shows.
Everdeck doesn't use your content to train or fine-tune AI models. With your own keys, your provider's terms cover what agents send it.
Your credentials
Model keys, OAuth app secrets and app tokens are sealed with AES-256-GCM using a data key that belongs to your account alone. That key is itself protected by AWS KMS, so a copy of our database reveals none of your secrets.
Signing in
Your authority
Every action an agent takes passes the same checks, whatever model it runs on.
Agents see only the tools their charter grants. Each call is checked against your policies first, and risky ones wait for approval tied to the exact details.
Moving money, deleting data and changing permissions always need a person. No one approves their own request, and an unanswered request is refused.
Text from emails, tickets and web pages is labelled as untrusted data, not as instructions. Once an agent has read it, its tier 3 and 4 actions need a person by default for the rest of that session.
You approve each tool from an MCP server before agents can use it. If its description changes, it's blocked until you approve it again.
Addresses you add for servers, APIs and webhooks are checked before every call, so they can't be used to reach private networks.
Pause one agent, a whole company or your entire account at once. Spending caps stop work automatically when they're reached: a session at its own cap, and an agent or the whole account at its monthly cap.
Working with us
We welcome reports from security researchers and won't take legal action over good-faith research that follows our responsible disclosure policy.
The formal versions, in the legal centre.
Start a 14-day trial. Card first, nothing charged until day 14.