Security is Not a Feature. It's the Foundation.
Your business data, your AI employee's conversations and your customers' details are protected by encryption, access control and a database that keeps every account apart.
Encryption
AES-256 encryption at rest and TLS in transit for every record, conversation and workflow. Keys to the tools you connect are encrypted a second time, at the application, with AES-256-GCM.
Infrastructure
Runs on Google Cloud Run, Google's managed platform, with Google Cloud SQL for data. Google patches the underlying servers; our services hold no persistent state of their own.
Access Control
Role-based access with owner, admin and user roles. Every account's data is kept apart at the database, with audit logging for admin actions.
Data handling
Your data is stored in the United States (Google Cloud, us-central1). Ask us to export or delete it at any time.
API Security
Short-lived sign-in tokens, rate limiting, input validation and strict cross-site rules on every request.
Your Business Stays Yours
Your AI employees, data and keys belong to your account alone. No other business can see them, and your AI employee can reach only the tools you connect.
Command oversight
Security for a business with an AI employee is also control over what it does. These are the controls, as they work today. How the parts fit together.
A log of every agent action
Every tool call an agent makes is written to a log with the agent, the record it touched, the outcome and the time. The log is what the daily report and the activity feed are built from. The arguments themselves are not stored, only a fingerprint.
A pause for any agent
An owner or admin can pause any agent from the console, for a set time or until resumed. A paused agent is not woken by its schedule, a reminder or a heartbeat and takes no new work. Work already in progress runs to its end.
On task, by design
Each kind of work carries a level, off, suggest, draft or autonomous, granted per task. Work with no grant is treated as draft; work set to off is refused and logged. Outbound texts, deletions and forms on other sites never pass draft. An agent with two incidents in a week is stepped back to draft.
Credentials
Keys to the tools you connect are encrypted with AES-256-GCM, stored as secrets the AI employee never sees in its instructions, and deleted when you remove the tool. Email and calendar are connected by OAuth, so we never hold your password, and you can revoke access from your Google or Microsoft account at any time.
The ledger, measured
Every tool call counted, every denial on the record.
How many times the workspace's agents called a tool in the window, as the tool-call ledger recorded it.
- Formula
SUM(calls) over GET /agents/tools/usage; calls = COUNT(*) FROM agent_tool_calls WHERE tenant_id = $tid AND created_at > now() - <days> GROUP BY tool_name (top 400 tools)- Sources
- agent_tool_callsGET /api/v1/agents/tools/usage
- League analogue
- Touches
- Value here
- — · computed live inside your workspace; this page carries no customer numbers
- Window
- The board's date selector
agent_tool_calls (one row per call, with its outcome) — not vertex_usage.tool_calls, which fleet.tool_calls sums; the two can differ by calls made outside a model run.
Regulated industries
Healthcare and dental
Front-office work only: hours, location and appointment requests. No patient care and no clinical records.
Legal
Intake only. An AI employee never gives legal advice.
Financial services
Inquiries and scheduling. Never investment or tax advice.
Have Security Questions?
Our security team is available to discuss your requirements, provide documentation, or schedule a security review.
Blue Heron Companies LLC · 14100 Walsingham Road, Largo, FL 33774
