Claude Managed Agents Goes Public Beta — Anthropic’s Enterprise Agent Runtime Is Here
Anthropic officially launched Claude Managed Agents in public beta today on the Claude Platform. The new cloud-hosted runtime handles secure execution, state management, tool orchestration, and long-running task scheduling for Claude agents — removing the need for enterprises to stand up their own infrastructure. Anthropic says the service cuts agent deployment time by up to 10x and supports both single-task flows and multi-agent pipelines.
Early adopters include Notion (workspace automation), Rakuten (finance ops), Asana (project management agents), Vibecode (code generation), and Sentry (incident triage). Managed Agents integrates with Claude Code, the Files API, MCP connectors, and the Message Batches API, and exposes observability hooks for teams that want to instrument agent behavior in production.
Accenture and Anthropic Announce Multi-Year Partnership to Scale Claude Across the Fortune 500
Accenture and Anthropic unveiled a multi-year partnership focused on moving enterprises from AI pilots to production deployments. Accenture will stand up a dedicated Claude practice, train 30,000 consultants on Claude’s agentic platform, and build pre-packaged industry accelerators for banking, insurance, healthcare, and manufacturing. The two companies will jointly invest in a “Center for Agentic AI” targeting Fortune 500 transformations.
The deal positions Accenture as the largest systems integrator working on Claude deployments and gives Anthropic a go-to-market motion it has historically lacked outside of cloud marketplaces. Initial client rollouts are underway at a handful of named Accenture accounts, with expanded rollouts expected throughout Q2 and Q3.
Snowflake Expands Anthropic Partnership With $200M Commitment
Snowflake and Anthropic announced a $200M expanded partnership that embeds Claude directly in the Snowflake platform and makes Claude available to more than 12,600 Snowflake customers. The expanded deal includes joint go-to-market across AWS, Azure, and Google Cloud, integration with Snowflake Cortex, and a shared roadmap for agentic workloads running on enterprise data warehouses.
The announcement lands on the same day as the Accenture partnership and the Managed Agents launch, hammering home Anthropic’s message: Claude is now a production-grade platform, not just an API. Enterprise revenue already makes up roughly 80% of Anthropic’s business per recent disclosures.
Claude Code Update: Larger MCP Result Persistence, Multi-Line Deep Links, Skill Shell Hardening
The latest Claude Code release expands MCP tool result persistence so larger payloads — think database schemas, log dumps, and full API responses — can pass through annotations up to 500K characters. The update also disables inline shell execution for skills and commands by default, closing a class of prompt-injection vectors flagged after last week’s security disclosures. Multi-line deep links are now supported, improving IDE round-trips.
Other quality-of-life changes: stronger session resume, better error messaging when tool calls exceed limits, and a new --mcp-result-limit flag for overriding the default truncation. The changes roll out automatically on the stable channel.
1M Token Context Window Beta Sunset for Sonnet 4.5 and Sonnet 4 — April 30 Deadline
Anthropic is reminding developers that the 1M token context window beta on Claude Sonnet 4.5 and Claude Sonnet 4 ends on April 30, 2026. To keep using 1M context, teams need to migrate to Claude Sonnet 4.6 or Claude Opus 4.6, both of which support the full 1M window at standard pricing with no beta header required. Anthropic also reminded teams that Sonnet 3.7 and Haiku 3.5 were retired earlier this month.
MCP Dev Summit: Anthropic, AWS, Microsoft, OpenAI Lay Out Enterprise Security Roadmap
At the MCP Dev Summit in New York earlier this week, maintainers from Anthropic, AWS, Microsoft, and OpenAI held a roundtable outlining the enterprise security roadmap for the Model Context Protocol. Priority areas include standardized auth flows for MCP servers, signed tool manifests, capability-based permissions, and a reference governance model for multi-tenant deployments. Official SDKs now see more than 97 million monthly downloads across Python and TypeScript — a reminder of just how fast MCP has become the de facto standard for agent tooling.
Claude Services Back to Green After Three-Day Outage Streak
Anthropic’s status page shows all Claude services operating normally on Thursday, the first clean day after a bruising three-day streak that included a ten-hour cascading failure on April 6, elevated API errors on April 7, and Sonnet 4.6 error spikes on April 8. Anthropic has not yet published a full post-mortem but told customers that additional load-balancing guardrails and a rollback of the configuration change that triggered Monday’s incident are already in place.
For users, the return to stability lands alongside a wave of new enterprise announcements — a reminder that the reliability story is now a narrative Anthropic has to manage, not just a technical problem to solve.
Plugin Marketplace Passes 2,500 Listings as Partner Plugins Keep Climbing
Community trackers now put the combined Claude Code plugin ecosystem at over 2,500 entries across the official Anthropic marketplace and community registries. The official marketplace carries 101 curated plugins (33 Anthropic-built, 68 partner), including entries from GitHub, Playwright, Supabase, Figma, Vercel, Linear, Sentry, Stripe, and Firebase. The long tail on community marketplaces has pushed past 2,400 skills and 770+ MCP servers.
Anthropic’s Platform Play Lands — And It’s Bigger Than Today’s Headlines
Managed Agents, Accenture, and Snowflake shipped on the same day for a reason. Anthropic is telling enterprises a single story: pick Claude, and you don’t have to build the plumbing, hire the integrators, or bolt it onto your data warehouse. That’s a direct shot at the “raw API + your team builds everything” posture that defined the first wave of frontier model deployments. It also mirrors the playbook AWS used to turn EC2 into a platform — managed services, certified partners, marketplace integrations.
The risk is obvious: every one of those layers is a dependency that has to stay up. Three straight days of outages earlier this week show how quickly a reliability wobble can reach every agent running on the platform. If Managed Agents becomes the default way enterprises deploy Claude, Anthropic’s SRE posture stops being a backend concern and becomes a board-level metric. The bet is that $30B in run-rate revenue buys the talent and hardware to get there before the next cascading failure.