Capability Discovery
How an agent finds out at runtime what tools and abilities are available to it.
Think of it like
Walking into a new workshop and reading the labels on the drawers before you start a job.
Example
On connecting to an MCP server, the agent asks “what can you do?” and gets back a live list of tools it can call.
How it actually works
Instead of hard-coding a fixed tool list, discovery lets an agent query its environment and adapt to whatever’s connected. This is a big part of why protocols like MCP took off — plug in a new server and the agent just sees the new tools. The catch is that more discoverable tools means more choices to get wrong, so discovery usually pairs with filtering to keep the active set manageable.
For product teams
Discovery is what lets one agent work across many tools without a code change for each — it’s the extensibility story.
For engineers
The agent queries connected servers or a registry at runtime to enumerate available tools and their schemas.
Related
- Tool Registry — The menu it discovers.
- MCP — The protocol that made dynamic discovery common.
Read anything AI without the jargon
Look up any term in plain English, or save terms as you read with the free Chrome extension.
Open DecoderAdd to Chrome