liveBook · Manning

Managed tool

As explained in Designing AI Systems

A managed tool is an externally invocable capability governed by a platform through registered identity, lifecycle, ownership, credentials, and operational controls.

A prototype can let an assistant call an external function directly; production action requires stronger control. Booking appointments, submitting insurance checks, and similar operations create consequences that multiply with scale, so tool access needs governed discovery, credentials, execution, and rate limits. section 6.0

A managed tool is a platform-governed capability rather than an application-local function. The platform assigns identity, tracks lifecycle and ownership, controls permissions, manages credentials, and executes the integration; applications declare the capabilities they need. section 6.1

A managed tool typically records:

  • A unique name, description, version history, owner, and access policy. section 6.1
  • Operational limits such as requests per session, with execution controlled centrally. section 6.0

Register once, discover wherever authorized

A scheduling integration can be registered as healthcare.scheduling.book_appointment with patient and slot parameters, a clinic API endpoint, non-read-only and non-idempotent behavior, and a three-request session limit. An authorized application can then discover tools in the healthcare.scheduling namespace and pass the returned definitions directly to model function calling.

How Designing AI Systems covers it

How deep it goes
Defines identity, lifecycle, ownership, permissions, registration, discovery, schemas, endpoints, behavior flags, and rate limits.
Examples
Uses healthcare appointment booking with patient and slot parameters, a scheduling endpoint, authorization, and a session request limit.
What it assumes
Assumes function calling and external integrations, while centralizing credentials, registration, discovery, and execution that applications otherwise implement themselves.
What you can do after
Apply it via an SDK to register integrations, discover authorized tools, and provide their definitions to model function calling.
How it connects
Connects tool discovery, function calling, credential management, access policies, rate limiting, and centralized security.

Related concepts

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