Anthropic adds session budgets, mid-session advisors, and data residency controls to Claude Managed Agents

Anthropic shipped four updates to Claude Managed Agents on August 7, giving developers hard spend caps, a way to have one model consult another mid-task, control over where inference runs, and GitHub-hosted skill loading.

What's new

Per Anthropic's Claude Platform release notes: "You can now set a budget on a Claude Managed Agents session: a hard cap on the session's spend, priced at public list rates. A session that reaches its budget pauses with the budget_reached stop reason instead of starting new model requests; changing or removing the budget resumes it." Deployments can apply the same budget to every session they start.

Sessions can also now be given an advisor — "a model at least as capable as the agent's own that the session's primary thread can consult mid-turn for strategic guidance," configured as an {"type": "advisor"} entry in the agent's multiagent roster.

Developers can additionally control where model inference physically runs for an agent by setting inference_geo inside the model configuration, either at agent-creation time or overridden per session, per Anthropic's data residency documentation. And Managed Agents sessions can now load skills directly from a mounted GitHub repository: any skills placed in a session's root .claude/skills directory are discovered automatically at session start.

Context

Claude Managed Agents is Anthropic's hosted infrastructure for running autonomous, multi-step Claude agents outside a developer's own orchestration code. The past several weeks of release notes show Anthropic rapidly filling out the product's operational controls — effort-level configuration on July 22, webhook coverage for environment and memory-store lifecycle events the same week, and now budget caps, advisor models, data residency, and repo-based skill loading.

Why it matters

Hard budget caps and configurable inference geography are the kind of guardrails enterprise buyers typically require before putting autonomous agents into production — uncontrolled spend and data-residency uncertainty are two of the most common blockers cited for adopting agentic AI at scale. By adding both in the same release, Anthropic is positioning Managed Agents less as a developer preview and more as infrastructure ready for regulated or cost-sensitive enterprise deployments.