Capability Matching

Capability matching is the process that allows Agistry agents (or LLM-based planners) to understand which adapters can fulfill a given intent — without requiring manual hardcoding. It powers dynamic tool selection, safe execution, and intelligent flow composition.

---

When a user prompt or trigger enters the system, Agistry evaluates which capabilities (adapters) are available and which ones match the input context and task.

How Capability Matching Works

Each adapter registered in Agistry includes structured metadata:

• name

• description

• inputSchema (required fields + types)

• tags

When an agent or planner runs, Agistry:

1. Parses the intent (natural language input)

2. Loads all agent-usable adapters from the registry

3. Passes adapter descriptions + schemas into the LLM context

4. The LLM returns a structured plan (tool name + input)

5. The runtime validates schema + executes in order

Capability matching lets Agistry go beyond rigid pipelines — enabling flows that adapt, agents that plan, and systems that evolve. By leveraging metadata, schema, and language context, Agistry transforms raw intent into executable plans — safely and intelligently.