Anthropic Adds Domain Controls and Cost Tracking to Claude Managed Agents
Summary
- • Anthropic shipped domain controls for web_search and web_fetch tools, letting teams restrict or block specific domains in Claude Managed Agent configs via allowed_domains or blocked_domains fields.
- • A redesigned Console session viewer adds cost tracking, a timeline minimap, and an Inspector panel showing raw events, per-tool stats, mounted resources, and per-thread activity.
- • Both updates target enterprise governance, making multi-agent sessions observable and web-tool access controllable for the first time on the platform.
- • Updates released August 19, 2026, building on the Claude Managed Agents platform launched earlier in 2026 and announced via official @ClaudeDevs posts.
Details
Domain controls for web_search and web_fetch tools
Teams can configure allowed_domains or blocked_domains on tool entries in the agent_toolset_20260401 configs array, restricting which websites Claude agents can access during web tool use — a core requirement for data-security-sensitive enterprise deployments.
Redesigned Console session viewer with cost tracking
An Inspector panel displays session details, cost, raw events, per-tool statistics, mounted resources, and per-thread activity for multi-agent sessions, giving developers a detailed financial and operational record of every agent run.
Timeline minimap and grouped transcripts added to Console
The Console viewer adds a timeline minimap and groups transcripts by model request, providing structured navigation for complex, multi-step agent sessions that span many interactions.
Updates extend Claude Managed Agents platform (2026)
The governance features build on the Claude Managed Agents platform launched earlier in 2026, adding control and observability layers to an already-deployed agent orchestration system, as documented in Anthropic's platform release notes.
Governance features target regulated-industry enterprise adoption
Domain allow/block lists and cost tracking are core audit and compliance requirements for enterprises in finance, healthcare, and legal — industries cautious about adopting AI agents due to data exfiltration and budget predictability concerns.
Details compiled via Grok live web research
Primary article text was unavailable; details sourced from AI Catchup coverage of the August 19, 2026 release and official @ClaudeDevs announcements referencing Claude platform release notes.
Anthropic Claude Managed Agents platform update, August 19, 2026. Sources: AI Catchup, @ClaudeDevs (official), Claude Platform Release Notes. Details supplemented via Grok live web research.
What This Means
Anthropic's addition of domain controls and cost tracking to Claude Managed Agents represents a meaningful step in making agentic AI deployable in regulated enterprise environments. Domain allow/block lists address the compliance concern that AI agents might access unauthorized resources or exfiltrate data; cost tracking gives finance and operations teams the budget predictability required for production rollouts. As long-horizon agent workflows move from demo to mission-critical, this kind of governance infrastructure is becoming table stakes — and Anthropic is building it ahead of broader enterprise adoption curves, directly challenging workflow-automation incumbents on their core enterprise selling points.