Security

Private by architecture

An assistant with access to your systems is only as safe as the architecture around it. Analyst starts from isolation: a dedicated environment per customer, least-privilege access to each tool, and approval gates on anything that changes the world.

One environment per customer

Your Analyst runs in a dedicated, isolated environment. No shared queues, no multi-tenant processing, no co-mingled data. Your credentials, your data, your Analyst.

Scoped, least-privilege access

Analyst connects to each tool with access scoped to the minimum a task needs. Grant exactly what you're comfortable with, and revoke any capability instantly.

Approval-first on anything mutative

Analyst reads and analyzes freely. Anything that sends, writes, or changes something waits for your explicit approval in the thread. You relax gates per workflow only as trust builds.

Your data trains no one

Everything Analyst learns about your team stays on your environment and improves only your Analyst. Nothing is pooled into a shared model or another customer's environment.

Full audit trail

Every read, action, model call, and approval is logged on your environment: what Analyst touched, which model handled which step, who approved what. Export anytime.

Clean offboarding

Leaving means your environment is destroyed: the running agent, its memory, logs, and learned routines, everything. We keep what's needed for invoicing and nothing else.

The environment

What runs where

A plain-language map of the architecture, because security pages that hide behind diagrams usually have something to hide.

Your tools (Slack, CRM, docs, billing, ...)
└─ Scoped, least-privilege access · nothing more
Your isolated environment (dedicated container)
├─ Your Analyst · reasoning and routines, running for you alone
├─ Model router · sends each step to the right model
├─ Memory · routines and preferences learned from your team
└─ Audit log · every read, action, and approval
Actions out
└─ Anything mutative is staged and waits for your approval
Model providers
└─ Receive only the minimum context per step, under no-training terms
Compliance

In progress, in the open

SOC 2 Type II is in progress. Until the report is in hand we won't put a badge on this page; we'll show you the architecture, the audit trail, and the controls instead.

Working with your security team

Bring your questionnaire

We complete security reviews as part of onboarding: questionnaires, architecture walkthroughs, and custom data-handling terms. If your team wants the environment inside your own cloud account, ask us about that.

Now onboarding early teams

Put an analyst on every team.

We connect your Slack, you show it one recurring task, and it runs that work for your team by the end of the first session.

Request accessSee how it works

Or write to hello@analystlabs.ai