Anthropic rolls out 10 Wall Street Claude agent templates and persistent 'Dreams' memory

Anthropic has spent the past two weeks expanding two related parts of its Claude product line: the infrastructure for long-running AI agents and the packaged workflows it is pitching to financial-services teams. On May 6, the company said Claude Managed Agents gained a research-preview memory feature called “dreaming,” plus outcomes, multiagent orchestration and webhooks. A day earlier, on May 5, Anthropic published a new set of 10 finance-focused agent templates for tasks such as pitch building, model maintenance and KYC review. On May 11, Anthropic added “agent view” in Claude Code, a separate research-preview dashboard for managing multiple coding sessions from one screen. ### What exactly did Anthropic ship in Managed Agents on May 6? Anthropic said on May 6 that “dreaming” is a scheduled process that reviews past agent sessions and memory stores, extracts patterns and curates memories so agents improve over time. The company described it as a research preview inside Claude Managed Agents on the Claude Platform, with access available by request. The same May 6 announcement said Anthropic also made outcomes, multiagent orchestration and webhooks available to developers building with Managed Agents. (claude.com) Outcomes lets a developer define a rubric for success and uses a separate grader to evaluate an agent’s output and trigger another pass if needed, Anthropic said. Anthropic said in internal testing that outcomes improved task success by up to 10 points over a standard prompting loop, including gains of 8.4% on docx generation and 10.1% on pptx generation. (claude.com) Those figures were presented by the company in its own benchmarks. ### How do Managed Agents differ from a standard model API call? Anthropic launched Claude Managed Agents in public beta on April 8 as a hosted system for running autonomous agents with built-in infrastructure. (claude.com) The company said the product is meant for long-running and asynchronous work, with secure containers, tool execution, checkpointing, credential management, scoped permissions and persisted event history handled by Anthropic’s platform. The platform documentation says developers create an agent, create an environment, start a session and then send events while Claude autonomously uses tools and streams results back. Anthropic positions that as an alternative to building a custom loop on top of the Messages API. ### What are the 10 Wall Street templates, and where do they fit? Anthropic said on May 5 that it was releasing 10 “ready-to-run” agent templates for financial services as plugins in Claude Cowork and Claude Code, and as cookbooks for Claude Managed Agents. (claude.com) The company said the templates target work including pitchbooks, KYC files and month-end close. The published list includes five templates for research and client coverage — Pitch builder, Meeting preparer, Earnings reviewer, Model builder and Market researcher — and five for finance and operations — Valuation reviewer, General ledger reconciler, Month-end closer, Statement auditor and KYC screener. (platform.claude.com) Anthropic said each template packages skills, connectors and subagents that firms can adapt to their own modeling conventions, risk policies and approval flows. (anthropic.com) Anthropic also said these finance updates pair with Claude Opus 4.7, which it said leads Vals AI’s Finance Agent benchmark at 64.37%. That performance claim came from Anthropic’s May 5 announcement. ### Where does Agent View fit into this rollout? Anthropic said on May 11 that agent view in Claude Code gives users “one place to manage all your Claude Code sessions.” The feature shows which sessions are waiting for input, still running or finished, and lets users start background sessions with `claude agents` or `claude --bg [task]`, according to Anthropic’s blog post and documentation. (anthropic.com) The Claude Code docs say agent view is in research preview and requires Claude Code version 2.1.139 or later. (anthropic.com) Anthropic said the feature is available on Pro, Max, Team, Enterprise and Claude API plans, with standard rate limits applying. ### What should users watch next? Anthropic’s May 6 and May 11 posts both frame key pieces of this rollout as research previews rather than general availability. The company’s next concrete step is continued access expansion through the Claude Platform and Claude Code documentation, where Managed Agents quickstart pages, agent-view setup instructions and the financial-services marketplace are already live. (claude.com 1) (claude.com 2)