Anthropic March 2026 release roundup

Anthropic rolled out a broad set of March 2026 updates across Claude Code, the Claude Developer Platform, Claude apps, and enterprise partnerships. Changes focused on larger context windows, workflow improvements, reliability fixes, visual output features, and new partner enablement programs.

Anthropic expanded Claude Code significantly in March 2026 with multiple releases focused on developer workflow, session control, and reliability. Version 2.1.76 added MCP elicitation support, new Elicitation and ElicitationResult hooks, a startup name flag, sparse worktree paths, a PostCompact hook, an /effort command, and a configurable session quality survey. That release also fixed a long list of issues affecting deferred tools, slash commands, plan approval, voice mode, model handling, Bash permissions, auto-compaction, MCP reconnection, LSP plugins, clipboard behavior, transcript scrolling, login flow, Remote Control, bridge session recovery, and hidden commands, while improving worktree startup performance, background agent behavior, model fallback visibility, dark-theme blockquotes, cleanup of stale worktrees, session titles, and voice-language messaging.

Claude Code 2.1.75 introduced Added 1M context window for Opus 4.6 by default for Max, Team, and Enterprise plans (previously required extra usage). It also added /color, session name display in the prompt bar, memory freshness timestamps, and hook source visibility in permission prompts. Fixes covered voice mode activation, model-name refresh after switching, attachment-related crashes, Bash handling of piped commands with !, plugin visibility, token estimation, marketplace config paths, resumed session names, Esc behavior in status dialogs, plan input handling, and footer hints. Improved startup performance on macOS non-MDM machines by skipping unnecessary subprocess spawns. A breaking change removed the deprecated Windows managed settings fallback at C:ProgramDataClaudeCodemanaged-settings.json and directed users to C:Program FilesClaudeCodemanaged-settings.json.

Earlier Claude Code releases carried more platform and compatibility changes. Version 2.1.74 added actionable /context suggestions and a configurable autoMemoryDirectory, while fixing memory leaks, managed policy bypasses, full model ID parsing, MCP OAuth issues, macOS microphone permissions, SessionEnd timing, plugin installation, marketplace submodule syncing, slash command handling, RTL rendering, Windows LSP file URIs, and VS Code session and scrolling issues. Version 2.1.73 added modelOverrides and SSL guidance, fixed freezes, deadlocks, lost Bash output, subagent model downgrades, background process cleanup, resume picker behavior, /ide crashes, /loop availability, double-firing SessionStart hooks, hook-generated no-op reminders, voice mode corruption, Linux startup and module issues, Remote Control image errors, and Windows heapdump behavior. It also changed the default Opus model on Bedrock, Vertex, and Microsoft Foundry to Opus 4.6.

Anthropic also updated the Claude Developer Platform. The 1M token context window is now generally available for Claude Opus 4.6 and Sonnet 4.6 at standard pricing. Requests over 200k tokens work automatically for these models with no beta header required. The 1M token context window remains in beta for Claude Sonnet 4.5 and Sonnet 4. We’ve removed the dedicated 1M rate limits for all supported models. We’ve raised the media limit from 100 to 600 images or PDF pages per request when using the 1M token context window.

Outside developer tooling, Anthropic added custom charts, diagrams, and inline visualizations to Claude and Claude Apps, and upgraded Excel and PowerPoint add-ins so they can share conversation context, support skills, and connect through an LLM gateway for Amazon Bedrock, Google Cloud’s Vertex Artificial Intelligence, and Microsoft Foundry users. On the enterprise side, Anthropic launched the Claude Partner Network and committed an initial 100 million to support training, technical support, market development, certifications, and co-marketing. The company also said it is scaling its partner-facing team fivefold and highlighted adoption efforts including training 30,000 Accenture professionals and supporting an organization of roughly 350,000 associates through partner activity.

65

Impact Score

Finance officials raise banking security concerns over Anthropic’s mythos model

Anthropic’s Claude Mythos has prompted urgent discussions among finance ministers, central bankers and banks over the risk that advanced cyber capabilities could expose weaknesses in critical financial systems. Governments and financial institutions are being given early access to test and strengthen defences before any broader release.

Uk delays Artificial Intelligence copyright reform

The UK government has postponed immediate copyright reform for Artificial Intelligence, leaving developers, creatives, and rightsholders to operate under existing law. Licensing, transparency, digital replicas, and future litigation are now set to shape the next phase of policy.

Memory architecture is central to autonomous llm agents

Memory design, not just model choice, determines whether autonomous agents can sustain context, learn from experience, and stay reliable over time. A practical framework centers on how information is written, managed, and read across multiple memory types.

Contact Us

Got questions? Use the form to contact us.

Contact Form

Clicking next sends a verification code to your email. After verifying, you can enter your message.