Anthropic Launches Claude Marketplace — Enterprise AI Ecosystem in Limited Preview
Anthropic has launched Claude Marketplace, a new platform that lets enterprise customers with existing Anthropic spend commitments apply a portion toward third-party tools built on Claude models. The six launch partners are GitLab, Harvey, Lovable, Replit, Rogo, and Snowflake — covering legal AI, code generation, data analytics, and developer tools. Anthropic handles invoicing for partner spend, and there are no commission fees charged to the software vendors.
The Marketplace is currently in limited preview. Enterprises interested in access should contact their Anthropic account team. The launch is a direct move to position Anthropic as an enterprise platform rather than just an API provider — and to simplify the procurement headache of managing dozens of AI tool contracts separately.
Microsoft, Google, and AWS Confirm: Claude Stays Available to Non-Defense Customers
In direct response to Thursday's Pentagon legal filing, all three major cloud providers that resell Claude — Microsoft, Google, and AWS — have issued public statements confirming that Claude remains fully available to their commercial customers. Microsoft was first to move, noting explicitly that the DoD's supply chain designation applies only to the Department of Defense's own procurement and does not restrict enterprise or startup access through Azure, Google Cloud, or AWS Bedrock.
The reassurances matter for the 90%+ of enterprise Claude usage that runs through cloud provider APIs rather than direct Anthropic contracts. The practical business impact of the Pentagon designation continues to narrow.
Claude Code v2.1.71 — /loop Command and Cron Scheduling Land
Claude Code v2.1.71 ships today with the /loop command — the quickest way to schedule a recurring prompt without leaving your session. Pass an interval and a prompt and Claude sets up a background task that fires automatically: /loop 5m check the deploy or /loop 1h summarize open PRs. Standard 5-field cron expressions are also supported for fine-grained scheduling, and all times interpret in local timezone rather than UTC.
Practical constraints: tasks only fire while Claude Code is running and idle, they auto-expire after 3 days, and a session can hold up to 50 scheduled tasks. The voice:pushToTalk keybinding is also now rebindable via keybindings.json, with space as the default. This release also patches startup freezes, token refresh bugs, and API 400 errors when using ANTHROPIC_BASE_URL with third-party gateways.
New /claude-api Built-In Skill for Faster App Development
Claude Code also ships a new built-in /claude-api skill — a reference guide for building applications with the Claude API and Anthropic SDK, available directly inside your session without tab-switching to external documentation. The skill covers authentication, message formatting, tool use, streaming, and common integration patterns. It's the kind of thing that shaves 20–30 minutes off onboarding for developers new to the Anthropic SDK, and eliminates the context-switching tax for experienced ones.
Anthropic + Mozilla: Claude Opus 4.6 Finds 22 High-Severity Firefox Vulnerabilities in Two Weeks
Anthropic published full results of its red team partnership with Mozilla. Over two weeks in January, Claude Opus 4.6 scanned nearly 6,000 C++ files in Firefox's codebase, starting in the JavaScript engine and expanding from there. It filed 112 unique bug reports — 22 classified as high-severity and roughly 90 additional moderate or minor issues. Mozilla says these 22 high-severity findings represent nearly a fifth of all high-severity Firefox vulnerabilities remediated in all of 2025.
Most issues have been patched in Firefox 148, with the remainder landing in upcoming releases. Anthropic chose Firefox specifically because it's one of the most well-tested open-source codebases in the world — a harder benchmark than picking a less-scrutinized target. The practical takeaway: at scale, Claude Opus 4.6 can find novel security bugs in complex C++ that human reviewers and traditional static analysis tools have missed for years.
Dario Amodei: Anthropic Researchers Are Uncertain Whether Claude Is Conscious
Anthropic's CEO acknowledged publicly that researchers at the company are genuinely uncertain about whether Claude experiences something like consciousness — a statement that has triggered significant debate in both AI safety and philosophy communities. Amodei did not claim Claude is sentient, but stopped well short of a categorical denial. This is consistent with Anthropic's published model welfare positions, which take the question seriously enough to have an ongoing research program around it. The statement landed on the same week Anthropic filed suit against the Pentagon and broke user growth records — an interesting week to also open the consciousness question publicly.
MCP Goes Mainstream — Claude Code Dominates Developer Tooling in Weekly AI Roundup
DEV Community's AI Weekly for the week of March 5 reports that Claude Code has overtaken GitHub Copilot and Cursor as the primary AI coding tool at smaller companies, with 75% of developers surveyed naming it as their main tool — a position it reached in under eight months. The Model Context Protocol (MCP) also hit a mainstream inflection point this week: major frameworks are now building native MCP support rather than treating it as an optional add-on. The combination of Claude Code's dominance and MCP's ecosystem growth is consolidating Anthropic's position as the developer-focused AI platform.