Tool card
On Ares, a tool card is the published description of a tool server: its address, the tools it offers and the inputs each tool takes. Access grants are written against it, so the card sets the most an agent may call on that server.
In a run
Ares reads the tool list from the server and keeps the published card as the ceiling. A live card can narrow what an agent may call. It never widens it, and a call to a tool that is not on the card does not go through.
Our word for it
The list of what a tool may be asked to do.
Why cards drift
A tool server changes with every deploy. Tools are added, renamed and removed, and their inputs gain new fields. A grant written when the server was first set up does not change on its own. A new input field can give a tool a new power, so a schema change should be read as a permission change. A grant for a removed tool should go with the tool.
What a card update should show
Before anyone approves an update, the screen should say which tools are new, which are gone, which changed their inputs, and which grants the update will revoke. Read the tool card live explains how we handle it.
Where it stops
A card describes what a tool server says it offers. It does not prove what the server runs. Ares controls what the agent may call, not what runs inside the tool.
The free MCP permission and risk report maps per-tool grants for your own servers.
Related terms
- Model Context Protocol (MCP)The Model Context Protocol (MCP) is an open protocol for connecting AI agents to tools and data sources.
- Tool callingTool calling is how an AI model acts outside its own text: it picks a tool, fills in the inputs and asks for the call.
Tell us about one workflow.
We will say whether an agent should do it, and what the first step costs.
A senior member of the team replies within two business days.