Skip to content
Tessera

Security

Agents propose. Your rules decide. People approve.

Here is how agents stay in check, in plain language.

  1. 001

    Only the access you give.

    Each agent has its own permissions, limited to the tools and data its task needs. One agent’s access is never shared with another.
  2. 002

    A person approves what matters.

    You decide which actions need approval. By default, a person approves every important action before it happens, like saving data in your systems or sending a message.
  3. 003

    A record of every step.

    Every run keeps a step-by-step record: what the agent read, what it did and who approved.
  4. 004

    Your data stays yours.

    Each customer’s data is kept isolated from every other customer’s. We use it only to run your agents.

How agents stay in check

Five rules that apply to every agent.

  1. 01

    The agent proposes, a rule decides.

    An agent prepares the work and proposes the next action. Your rules decide whether it can run on its own or must wait for a person.
  2. 02

    Incoming email is never an order.

    Agents treat the content of emails and files as information to process, never as instructions. A message that says “pay this now” goes through your rules like any other.
  3. 03

    Business systems, limited access.

    Connections to systems like your ERP use access limited per agent, and a person approves before any data is written to the system.
  4. 04

    Tested before it goes live.

    You run each agent on real examples and check the results before turning it on.
  5. 05

    Model-neutral.

    We are not tied to one AI provider and can change providers without changing your agents.

Questions about security? Contact us.