Skills, tools & connectors
When an agent says it has checked your calendar, it isn't bluffing — somewhere in the last few seconds it actually opened your calendar and read it. The things an agent can actually do are called skills (you'll also see the word tools for the individual actions inside them): read a mailbox, search the web, create a JIRA issue, fill a spreadsheet, post a message. An agent without skills can only talk. An agent with the right skills can work.
Every agent carries its own set, switched on or off individually. That's why the CRM Assistant can update a deal but can't touch the file store, and the reporting agent can read project data but can't send email. It isn't a malfunction when an agent politely declines to do something a different agent did yesterday — it's the platform doing exactly what your admins asked of it. Capability follows role, the same way it does with people.
Two ways in: the organisation's connections, and yours
For a skill to reach a real system, something has to be plugged in, and EnClaw draws a careful line between two kinds of plug.
Org integrations are connected once by your admin and shared by everyone: the Slack workspace, Microsoft Teams, the CRM, the project tracker. When an agent files a ticket through an org integration, it acts on the organisation's behalf, under the organisation's rules.
Personal connectors are yours. When you connect your own Outlook, Gmail, JIRA or GitHub account, you're granting delegated access — the agent acts as you, sees only what you can see, and does it only for you. Your inbox does not become visible to your colleagues or their agents; a colleague asking an agent about "the inbox" gets their own, or nothing. You choose exactly which permissions to grant when you connect — read calendar but not email, say — and each permission is marked with how much trust it involves. Connections age too: a small badge tells you when one has expired and needs a quick re-authorisation, which is the cause of roughly half of all "the agent suddenly can't see my calendar" mysteries.
Where new abilities come from
Skills aren't a fixed menu. The marketplace offers ready-made ones — and ready-made agents, and whole squads — that can be installed into your workspace, and organisations can build their own custom skills for the work nobody else does, from generating branded documents to talking to an in-house system. Anything custom passes through a security review before an agent is allowed to use it; by the time a skill appears on an agent near you, it has been checked, approved and switched on deliberately.
So agents can act, and their reach is deliberate and bounded. The next question is subtler: what do they keep? You told an agent last month that you prefer short answers, and it still remembers. How that works — and who else can see it — is the next chapter.