Anthropic's Major Releases in 2026: Claude 4.6, MCP, Agent SDK & More

The first quarter of 2026 has been Anthropic's most productive period to date. New model families, a fully overhauled developer platform, a $30 billion funding round, and the open-sourcing of the MCP standard to a neutral foundation — all in under 90 days. Here is a structured overview of everything that shipped, and what it means for businesses integrating Claude today.

New Models: Claude Opus 4.6 and Sonnet 4.6

On February 5, Anthropic released Claude Opus 4.6 — the most capable model in the Claude 4 family. It introduces a 1-million-token context window and a new adaptive thinking mode that dynamically allocates reasoning depth to match task complexity. Pricing is $5 per million input tokens and $25 per million output tokens via the API.

On February 17, Claude Sonnet 4.6 became the new default model across Claude.ai and the API. It delivers near-Opus-4.6 performance at a significantly lower cost: $3/M input and $15/M output. Sonnet 4.6 is optimised for speed and throughput, making it the right choice for most production deployments.

A third model, internally code-named Claude Mythos, was briefly leaked on March 26 but has not yet been officially announced. Early signals point to a multimodal leap beyond current benchmarks.

Claude Code: Computer Use, Auto Mode & Agent Teams

Claude Code received several landmark updates in Q1 2026. On March 23, computer use went generally available — Claude can now autonomously control a desktop environment, navigate browsers, fill forms, and execute multi-step workflows that were previously only possible with a human operator.

On March 24, Anthropic shipped Auto mode, which lets Claude Code intelligently switch between tools and reasoning strategies without manual orchestration. Combined with a new PowerShell tool for Windows environments and native Xcode integration for iOS/macOS developers, the platform now covers all major development environments.

Agent teams — the ability to have multiple Claude instances collaborate on a task with specialised roles — entered private beta in March. This capability, combined with the 1M context window of Opus 4.6, unlocks genuinely new use cases: full codebase reviews, enterprise-scale document analysis, and end-to-end process automation without human checkpoints.

MCP Donated to the Linux Foundation

The Model Context Protocol (MCP) — Anthropic's open standard for connecting AI agents to external tools and data sources — passed 97 million installations by Q1 2026. On March 11, Anthropic donated MCP to the newly formed Agentic AI Foundation under the Linux Foundation umbrella, ensuring neutral governance and multi-vendor participation.

This move signals that MCP is becoming the de facto industry standard for agent interoperability, much as HTTP became the standard for web communication. With 75+ enterprise connectors now generally available (Salesforce, SAP, Microsoft 365, Google Workspace, Jira, Confluence, and dozens more), the integration story for enterprise teams has never been clearer.

Agent SDK: The Successor to Claude Code SDK

The Claude Agent SDK — the renamed and expanded version of the Claude Code SDK — reached GA in Q1 2026. It provides structured primitives for building multi-agent pipelines in both Python and TypeScript, with first-class support for tool use, memory, and long-running tasks. The SDK now ships with built-in support for MCP connectors, dramatically reducing boilerplate for enterprise integrations.

Claude for Work: Cowork & Enterprise Features

In January, Anthropic launched Claude Cowork — a shared workspace for teams — initially on the Max plan (Jan 12) before expanding to the Pro plan (Jan 16). Cowork enables collaborative AI sessions, shared prompts, and team memory, bridging the gap between individual productivity and organisation-wide AI adoption.

Computer use landed in Cowork on March 23, meaning enterprise teams can now delegate browser-based workflows to Claude directly from their shared workspace — no custom infrastructure required.

Partner Network: $100M Committed, Big 4 On Board

On March 12, Anthropic announced a $100 million commitment to its partner network, with anchor partners including Accenture, Deloitte, Cognizant, and Infosys. These partnerships expand Claude's reach into large enterprise deployments across finance, healthcare, logistics, and professional services.

Alongside the investment, Anthropic launched the Claude Certified Architect professional certification — a structured exam-based qualification for developers and consultants building production AI systems on Claude. AI Workshop by Blocksize is a certified Anthropic partner and eligible trainer for this certification path.

Safety & Policy: RSP v3.0 and the Anthropic Institute

Anthropic published Responsible Scaling Policy v3.0 on February 24. The updated policy removes the previous "hard pause" commitment — a clause that required halting development if certain capability thresholds were crossed — replacing it with more nuanced evaluation-and-mitigation procedures. The change has drawn mixed reactions from the safety research community.

On March 11, Anthropic founded the Anthropic Institute, a dedicated internal research body focused on interpretability, alignment, and long-term AI safety. This move signals a structural commitment to safety research at scale, separate from the commercial product organisation.

Series G: $30 Billion at a $380 Billion Valuation

Anthropic closed its Series G funding round raising $30 billion at a $380 billion post-money valuation — making it one of the most highly valued private technology companies in history. The round provides a multi-year runway for frontier model research, safety work, and global infrastructure expansion.

What This Means for Your Business

Taken together, Q1 2026 represents a step-change in what AI can do for enterprises. The combination of Opus 4.6's million-token context, Claude Code's computer use capability, and MCP's mature connector ecosystem means that any knowledge-work process can now be meaningfully automated — not as a prototype, but in production.

For companies in Croatia and the DACH region considering Claude integration, the window to establish AI-driven competitive advantage is open right now. The tools are enterprise-grade, the pricing is predictable, and the partner ecosystem — including certified consultants like AI Workshop — can take you from pilot to production in weeks, not months.