AI capability glossary
Plain-English explainers for the concepts that show up across the AgentMaps catalog — MCP servers, skills, agent templates, verification levels, permission scopes, and the terms we use to describe risk. Each entry is bilingual and links back to the capabilities in the catalog it applies to.
MCP server
A local or remote program that exposes tools, resources, or prompts to an AI assistant using the Model Context Protocol (MCP) — the open standard for connecting AI apps to outside systems.
Agent template
A pre-built starting point for an AI agent — the prompt, tool wiring, and often the runtime loop — that you clone, configure, and adapt instead of building from scratch.
AI skill
A packaged unit of instructions — and optionally code — that an AI assistant loads on demand to do one specific task well. Think of it as a plugin the AI itself decides when to use.
Workflow template
A pre-built multi-step pipeline — often chaining tool calls, API requests, and conditional branches — that you fork and configure rather than build from scratch.
Verification level
How far a capability has been checked on AgentMaps — from just registered (Listed) to interface-tested (Interface-verified). It shows evidence depth, not production-readiness.
Risk level
AgentMaps' summary of what a capability could do to your system — low, medium, or high — based on its permission scope, sensitive-data access, and destructive-action potential.
Permission scope
The concrete set of things a capability can do — read your files, run shell commands, hit the network, write to a database — spelled out per capability rather than hidden behind a single "trusted" flag.
Shell access
A capability with shell access can run arbitrary commands on the host — install packages, delete files, launch processes. The most powerful and highest-risk permission an AI tool can hold.
Write access
A capability with write access can create or modify files on your disk. Less dangerous than shell access, but still capable of overwriting real work — so the write scope matters.
Sensitive token
A credential — API key, OAuth token, or service account — that a capability needs to authenticate on your behalf. Powerful, revocable, and worth scoping tightly.