What is Endpoint AI and when should we use it?

Endpoint AI (the Harriet capability you may see alongside Provisioner in contracts and admin URLs) is Harriet’s deployment and control surface for organizations that want Harriet beyond a single browser session—managed desktops, local assistants using Harriet tooling, or environments where IT needs consistent skill, connector, and tool policy.

When to use it

• You roll Harriet out to many machines or specific cohorts with the same configuration.
• You want auditability and governance over which external tools or skills appear on endpoints.
• Your organization uses Endpoint AI / Provisioner features and needs a clear owner for endpoint policy and review workflows for skills and MCP connectors.

When you might not need it

• Pilot teams using only web chat and Slack or Teams with standard admin controls may rely on Company settings alone until you need tighter endpoint management or org-wide review for packaged skills and connectors.

Example

IT deploys Harriet on consultants’ laptops with a fixed skill bundle and holds new MCP connectors in draft until security approves them in the review queue.

Guardrails

• Endpoint AI does not replace identity or SSO policy—it sits alongside them.
• Coordinate with IT on who may enrol devices, rotate API keys, and approve items in the review queue.

For packaged skills built locally with Skilify, submitting into Harriet, and how admins provision approved skills to teams, see Using Skilify to build skills locally, submit for review, and provision them.