Tools

Opened from the account menu (bottom-left) under Setup, or from ⌘K.

Browse every tool available on the deployment, grouped by category, each with a description of what it does. Tools that exist but are switched off show as locked with the reason — for example a tool that needs a credential you have not set yet — so you can find out a capability exists instead of it being invisible.

You do not have to come here to give an agent a toolkit. Ask the agent — "connect yourself to GitHub" — and it adds the toolkit to itself (self_add_toolkit), tells you which sign-in it needs and sends the link for it, then restarts when you say so; self_remove_toolkit takes one away again. AgentFather does the same for any agent (olano_add_toolkit). A toolkit added this way is the same as one you tick on the agent's Tools tab: it applies after the agent restarts.

Fleet-wide tools. A few tools are badged fleet-wide, with a note under the row explaining what that means. They act across the whole deployment rather than only for the agent holding them: reading the shared runtime log, another agent's chat transcripts or workspace, the credentials vault, or running a command as a different agent. Anyone who can chat with an agent can ask it to use the tools it holds — so a fleet-wide tool belongs only on an operations agent you trust with the entire deployment. AgentFather ships with them; nothing else gets them unless you tick them yourself.

Two rules back the badge up. Agents cannot hand these to each other or to themselves — a bot asked to "give yourself the log reader" is refused, and so is AgentFather doing it on a scheduled run or on another agent's behalf; adding one is a decision an administrator makes here, in the dashboard. And holding the tool is not enough to use it: these tools answer only when an administrator is the person on the turn, so an ordinary member chatting with an agent that holds them still cannot read another agent's logs or transcripts. Ordinary tools are unaffected — the badge is the whole list.