# Guardrails

Source: https://customlabs.io/glossary/guardrails/
Updated: 2026-09-13

Agents

# Guardrails

Guardrails are the checks that keep an LLM or agent inside acceptable bounds in production.

They include content filters, schema validation, permission scoping, and human-approval gates.

They matter most for agents with real tool access, where a bad response becomes an unwanted action.

Guardrails are a design layer added deliberately, not a property models come with by default.

[Capabilities](https://customlabs.io/capabilities/)

[← Back to the full glossary](https://customlabs.io/glossary/)

## Related terms

[Agent (Agentic AI) An agent is an LLM given a loop, memory, and tools it can call.](https://customlabs.io/glossary/agent/)[Human Oversight Human oversight is a checkpoint where a person can meaningfully approve, reject, or intervene.](https://customlabs.io/glossary/human-oversight/)[Prompt Injection Prompt injection is untrusted input crafted to override a model's system prompt or task.](https://customlabs.io/glossary/prompt-injection/)[Red Teaming Red teaming is deliberately attacking your own AI system to find what breaks first.](https://customlabs.io/glossary/red-teaming/)

## More in Agents

[Tool Calling Tool calling lets a model request a structured action instead of only generating text.](https://customlabs.io/glossary/tool-calling/)[Model Context Protocol (MCP) MCP is an open standard for connecting LLM applications to tools and data sources.](https://customlabs.io/glossary/model-context-protocol/)[Context Engineering Context engineering decides what occupies a model's context window at every step, and in what order.](https://customlabs.io/glossary/context-engineering/)[Tool Contract A tool contract is the schema a tool exposes to a model.](https://customlabs.io/glossary/tool-contract/)