Framer rebuilt the harness behind its Framer Agent, the layer that runs the agent's steps, so it can take on a whole workflow as a single program. It reads your document, makes decisions, and applies changes while the model is still writing the next steps, so you see progress sooner.
Framer says the new approach uses 2x fewer tool steps and 39% fewer cached input tokens per message. Those are the publisher's figures. The same capabilities are also available to external agents through MCP.
