AI developer tools · What shipped

Anthropic released Claude Fable 5.1 and Claude Mythos 5.1, advanced AI models for coding and knowledge work. Fable 5.1 achieves 25% lower costs for typical workloads and up to 45% savings for agentic tasks due to 75% reduced cache read pricing. The models show significant benchmark improvements: Fable 5.1 scores 52.6% on Terminal-Bench-Science 0.1 (agentic scientific research), 55.8% on Terminal-Bench 4.0 (agentic coding), 77.9% on OSWorld 2.0 (strict computer use), and 73.4% on CursorBench 3.2.0, outperforming Fable 5 across multiple tasks. Mythos 5.1, with enhanced safeguards for cybersecurity and life sciences, demonstrated protein design capabilities producing binding affinities 10x higher than competition entries and 50% hit rates on protein design tasks. Scientific applications include a new Venus elevation map and GPU kernel optimizations for genomic models achieving 2.5x speedups and 30-60% cost reductions on genome-wide analyses.

Agent Package Manager (APM), a dependency manager for AI agents comparable to npm, has shipped from Microsoft as an open source MIT-licensed project. APM allows developers to declare skills, prompts, instructions, plugins, and MCP servers in a single apm.yml manifest file, with support across multiple AI development environments including GitHub Copilot, Claude Code, Cursor, Gemini, and others. The tool includes a lockfile for reproducible deployments, security scanning for hidden Unicode and transitive MCP servers, and policy governance enforced at install time through apm-policy.yml with tighten-only inheritance across enterprise, organization, and repository levels.

Flowise, an open-source visual builder for LLM applications written in Node.js on LangChain.js, offers three builder types for increasing complexity: Assistant for simple RAG chat, Chatflow for single-agent systems, and Agentflow for multi-agent orchestration. The platform supports over 100 integrations with models, vector databases, and data sources, includes MCP client and server nodes for tool interoperability, and exposes finished flows as REST APIs with embedded chat widgets or via JS and Python SDKs. Each flow becomes an instantly deployable endpoint with execution logs, visual step debugging, input moderation, rate limiting, RBAC, SSO, and self-hosting capabilities to keep credentials and conversation data private.