liveBook · Manning

Tool registry

As explained in AI Governance

A tool registry is a centralized catalog of approved tools and the permissions governing their use.

A tool registry is an IT-controlled catalog of vetted tools that centralizes which integrations agents may access. Without centralized oversight, unreviewed integrations can multiply, while an agent compromised through prompt injection may inherit every capability exposed by its connected tools. section 4.8.4

Use the registry to make tool access and data handling more visible:

  • Catalog callable tools and the data each receives; execution traces record which tools agents actually invoke. section 5.6.4
  • Review the capabilities exposed by each integration: the cited GitHub server includes destructive operations, while Snowflake’s generic SQL tool accepts arbitrary queries. section 4.8.4

Tracing a data flow

An agent may combine CRM data with public web data, then send an output through an email tool. Documenting accessible tools and their constraints, alongside execution traces, helps track designed capabilities and actual use. section 5.6.4

How AI Governance covers it

How deep it goes
Defines a vetted catalog and links it to permissions, data-flow records, and execution traces, without detailing registry implementation.
Examples
Uses GitHub MCP destructive operations, Snowflake arbitrary SQL, and a CRM-to-web-to-email flow to illustrate exposure and data handling.
What you can do after
Apply a centralized catalog to document callable tools and their data access, then compare designed capabilities with execution traces.
How it connects
Connects agent permissions with MCP servers, prompt injection, data flows, execution traces, and GDPR records of processing activities.

Also explained in

Compare how 4 books explain tool registry →

Related concepts

This explainer was generated by Manning's knowledge graph from the text of AI Governance. The section links point to the parts of the book it draws on.