Claude Cowork, now cleared
for your most important work.

Tenet provides the missing compliance layer regulated industries require — so doctors, brokers, and advisors can use AI without worry.

Before

Claude Cowork is powerful. In regulated industries, it can also be a liability.

This is Claude out of the box — no enterprise controls, no compliance configuration, no guardrails.

PHI Exposure

Your patients' data enters the model. Claude Cowork reads files, emails, and notes to do its job. HIPAA doesn't care that it was helpful — PHI in a model context window is a disclosure event.

No Audit Trail

No one knows what the agent did. When an AI edits a document, sends a draft, or reads a record — there's no audit trail. Your compliance team can't reconstruct it. Neither can you.

No Scoping

The agent has no lane. Out of the box, Claude Cowork can read and act on anything you can. Insurance underwriting files. Patient histories. Client portfolios. It doesn't know what's off-limits.

Five non-negotiables when deploying AI today:

01

PII/PHI detection

Sensitive data must be identified and redacted before it reaches the model context

02

Immutable audit logs

Every agent action logged: what was read, what was written, what was sent, and when

03

Scoped file permissions

The agent operates only within boundaries you define

04

Human-in-the-loop gates

High-stakes actions require explicit approval before execution

05

Incident replay

When something goes wrong, you can reconstruct exactly what happened

Get started in minutes.

Drag Tenet to Applications 01

Download & Install

Download the macOS app and run the installer. The Tenet MCP server will start automatically.

Upload plugin in Claude Cowork 02

Connect to Claude Cowork

Add Tenet as a Custom Connector to Claude. Claude will automatically detect new skills.

Tenet scanning and redacting PII in Claude Cowork 03

Get to work, worry-free

Every Cowork action now routes through Tenet's policy engine: PHI stripped, actions logged, permissions enforced.

Research preview available now.

Tenet is available as a research preview for macOS. Download, install, and connect to Claude Cowork in under five minutes.

Frequently asked questions

Today, yes — Tenet is built specifically for Claude Cowork. ChatGPT works differently: your message goes to OpenAI’s servers before any privacy tool can act on it, so there’s no way to guarantee PII never reaches the model. More AI clients are on the roadmap.

No PII ever leaves your device — detection and redaction happen locally, and the model only sees sanitized placeholders. Pro and Enterprise deployments use cloud-based audit logging and compliance reporting. Tenet also collects optional, anonymous telemetry to improve the product, which can be disabled at any time.

Yes, you can do it yourself. Download the app, install it, and connect it to Claude Cowork — most users are up and running in under five minutes with no technical setup required. Enterprise deployments are different: organization-wide policy enforcement, admin controls, and BAA execution are handled with dedicated support from the Tenet team.

Tenet’s core features are free during the research preview and will continue to be opensource. A Pro plan is coming soon for individuals and small teams who need advanced controls. Enterprise pricing starts at $2 per 1M tokens, with a BAA, organization-wide policy enforcement, and dedicated compliance support included. Reach out to team@tenetlabs.com to discuss Enterprise.

Tenet for Enterprises

Healthcare providers, insurers, and financial institutions deploy Tenet with enterprise controls, a BAA, and dedicated compliance support.

  • Business Associate Agreement included
  • Organization-wide policy enforcement
  • Audit-ready logs and compliance reports
  • Multi-user deployment with admin controls
  • Dedicated compliance and security support
Let's talk about Enterprise controls →

Security & Data Handling

Tenet always runs entirely on your machine. Nothing leaves your device without first passing through the detection pipeline — out of the box, no configuration required.

  • PII/PHI redacted locally before reaching Anthropic's API — the model only sees sanitized placeholders
  • Original values stored on-device in encrypted local storage and rehydrated into final output
  • Full audit trail written locally — every redaction, every agent interaction logged
  • Works out of the box — default configuration meets research preview security standards

Enterprise deployments support custom security reviews, architecture documentation, and BAA execution. Research preview users are covered by local-only defaults.