AI IntelligenceAI-AGENT.NEWS

AI Engineering Topics

Tools and MCP

Tools let agents take bounded actions or retrieve information; MCP standardises how many tools and resources are exposed to clients.

833 matching public evidence items

All Topics · Capabilities and Interfaces

How it works

The runtime publishes schemas and permissions, the model requests a capability, and the application validates, executes and returns a structured result.

  1. Describe capability
  2. Select tool
  3. Validate request
  4. Return result

Key concepts

Tool schema
The typed contract for an action and its inputs.
MCP
A protocol for connecting models to tools, resources and prompts.
Capability boundary
The exact actions a tool is permitted to perform.

Engineering risks

  • Over-broad tools enlarge the impact of model mistakes.
  • Tool output can carry untrusted instructions.
  • Schema validation does not replace authorisation.

Aggregated from the site

Evidence and developments

Matching public material uses the same topic rules as AI Engineering.

public GitHub discoveries

rse/ase