Anthropic as we speak launched Claude Managed Brokers as a analysis preview. Dreaming extends reminiscence by reviewing previous periods to seek out patterns and assist brokers self-improve. We’re additionally making outcomes, multiagent orchestration, and webhooks out there to builders constructing with Managed Brokers. Collectively, these updates make brokers extra succesful at dealing with advanced duties with minimal steering.
Construct self-improving brokers with dreaming
Dreaming is a scheduled course of that evaluations your agent periods and reminiscence shops, extracts patterns, and curates reminiscences so your brokers enhance over time. You determine how a lot management you need: dreaming can replace reminiscence robotically, or you’ll be able to overview adjustments earlier than they land.
Dreaming surfaces patterns {that a} single agent can’t see by itself, together with recurring errors, workflows that brokers converge on, and preferences shared throughout a crew. It additionally restructures reminiscence so it stays high-signal because it evolves. That is particularly helpful for long-running work and multiagent orchestration.
Collectively, reminiscence and dreaming type a sturdy reminiscence system for self-improving brokers. Reminiscence lets every agent seize what it learns as it really works. Dreaming refines that reminiscence between periods, pulling shared learnings throughout brokers and conserving it up-to-date.
Dreaming is accessible in Managed Brokers on the Claude Platform; builders can request entry right here.
Ship higher outcomes
With outcomes, you write a rubric describing what success appears to be like like and the agent works towards it. A separate grader evaluates the output towards your standards in its personal context window, so it isn’t influenced by the agent’s reasoning. When one thing isn’t proper, the grader pinpoints what wants to vary and the agent takes one other go.
Brokers do their greatest work after they know what “good” appears to be like like. For instance, a structural framework, a presentation normal, or a set of necessities that have to be met. With outcomes, brokers can verify their work towards that bar and self-correct till the output is sweet sufficient, with out a human needing to overview every try.
Outcomes is especially helpful for duties that require consideration to element and exhaustive protection. It additionally works for subjective high quality, like whether or not copy matches a model voice or a design follows visible tips. In testing, outcomes improved job success by as much as 10 factors over an ordinary prompting loop, with the most important positive aspects on the toughest issues. Outcomes additionally improved file era high quality, with +8.4% job success on docx and +10.1% on pptx in our inside benchmarks.
You may as well now outline an final result, let the agent run, and get notified by a webhook when it’s carried out.
Deal with advanced duties with a number of brokers
When there may be an excessive amount of work for a single agent to do nicely, multiagent orchestration lets a lead agent break the job into items and delegate each to a specialist with its personal mannequin, immediate, and instruments. For instance, a lead agent can run an investigation whereas subagents fan out by deploy historical past, error logs, metrics, and help tickets.
These specialists work in parallel on a shared filesystem and contribute to the lead agent’s general context. The lead agent can verify again in with different brokers mid-workflow as a result of occasions are persistent and each agent remembers what it’s carried out. You may as well hint each step within the Claude Console: which agent did what, in what order, and why, providing you with full visibility into how your job was delegated and executed.
