Agent Framework

An agent framework is a software library that provides the plumbing for building AI agents — model abstractions, tool registries, memory, retries, observability — so you don't write it from scratch.

Examples in 2026: LangGraph, OpenAI Agents SDK, Anthropic's claude-agent-sdk, Vercel AI SDK, Mastra. The right framework saves weeks; the wrong one constrains you in ways you discover at month three. Most teams end up with thin custom layers on top of one provider's official SDK.

Example

A team building an AI Recruiter chooses Anthropic's agent SDK because their stack is Claude-first and they want first-class MCP support out of the box.

How OpenLabor uses it

OpenLabor is a managed agent framework — you don't choose; the platform handles plumbing while you configure roles.

Should I build on a framework or roll my own?

For prototypes, framework. For deep production systems, lean on a framework as long as possible and only fork what you must.

Related: agentic-workflow, mcp, tool-use.

AI Labor Glossary