What is an Agent Framework?

At its core, an agent is a loop: send context to a model, get back either a final answer or a request to call a tool, execute that tool, feed the result back in, and repeat until the task is done. A framework standardizes that loop along with the surrounding concerns — how tool definitions are declared, how conversation memory is stored and retrieved, how errors and retries are handled, and how permission checks are enforced before a tool call actually executes. Frameworks vary in how opinionated they are: some are thin libraries that just standardize the model-and-tool loop, while others are full platforms that also handle multi-agent coordination, sandboxed execution environments, and observability out of the box. The right choice depends on whether a team needs to move fast within a narrow use case or needs the flexibility to support many different agent types across an evolving product.