AGISTRY DOCUMENTATION
  • Overview
    • Introduction
  • How Agistry works?
  • Roadmap
  • agistry hub
  • Dashboard
  • Create your first Agent
  • Manage your Agents
  • Triggers
  • Adapters & Modules
  • Workflows
  • API Keys
  • Enterprise
  • Framework
    • Core Concepts
    • Getting Started
    • Building and Deploying
  • Execution Model
    • SDK
      • Authentication
      • Adapter Calls
      • Design Flow
    • Adapters
      • Adapter Specification
      • Adapter Schema
      • Adapter Execution
  • Registry
    • Registry Design
    • Capability Matching
  • official links
    • Website
    • GitHub
  • X/Twitter
  • Telegram
Powered by GitBook
On this page
Export as PDF
  1. Execution Model
  2. Adapters

Adapter Specification

Adapters are the atomic units of execution in Agistry — self-contained, intent-driven, and remotely orchestrated. They’re not just functions — they’re protocol-compliant components that define what should happen, what they need, and how they transform context inside a flow.

What Makes an Adapter?

Every adapter in Agistry is defined by:

  • A unique name (pdf-text-extractor, telegram-alert)

  • A description (used in UI & LLM prompt context)

  • An input schema (JSON structure it expects)

  • An output contract (optional, used for planning)

  • A runtime implementation (hosted by you or Agistry)

  • Optional visibility, versioning, and metadata

Adapters are typed, tracked, and composable. They can be chained, scheduled, triggered, or run recursively — depending on how your flows are designed.

PreviousAdaptersNextAdapter Schema

Last updated 17 days ago