vibehacker
News

MCP TypeScript SDK 2.4 adds the Skills extension, so servers can serve SKILL.md folders to agents

Released Oct 23, @modelcontextprotocol/server/ext/skills lets a server load a skills folder with loadSkills(folder) and serve it via skills/list, skills/get and skill:// resources, while the matching SkillsClient verifies each file's size, digest and frontmatter. Heads-up: x-mcp-header on number-typed tool params is now rejected per the 2026-07-28 spec, so switch those to integers.

More news

View all

Google ADK Python 2.13 promotes 12 features to stable and switches telemetry to OpenTelemetry GenAI v2

Released Oct 22, it adds TypesafeClassifier and ClassifierNode for routing workflows without an LLM call, live audio with barge in, and experimental MongoDB session and keyword match memory services that need no vector index. Heads up: telemetry now defaults to OTel GenAI v2 and drops the call llm span, so set ADK TELEMETRY SCHEMA VERSION OPT IN=otel semconv 1 36 if your dashboards depend on the old schema…

Chroma is joining Replit, and Chroma Cloud stops taking writes on Nov 22

Announced Oct 23, the ChromaDB team signed a definitive agreement to join Replit to build a context and memory layer for agents, and open source ChromaDB stays Apache 2.0 with more code moved in from private repos. Chroma Cloud is winding down: writes stop November 22 and reads stop December 22, 2026, so hosted users need to migrate (Magica)…

Chroma Blog

Beads 1.3.2 fixes two data-loss paths in bd purge and auto-backup for agent issue tracking

Released Oct 24, the fix only patch to Steve Yegge's agent memory and issue tracker stops bd purge retention sweeps from deleting a closed root that still has live work beneath it, and stops auto backup from skipping commits that land mid sync. It's a binary swap from 1.3.1 with no schema migration, and also lets least privilege Dolt users write again and unsticks bd export all on large wisp sets…

Hugging Face hands OpenEnv, its framework for agent RL training environments, to the AAIF

Announced Oct 22, OpenEnv packages RL environments as containerized services that expose an agent interface, MCP, and the Gym protocol training libraries use, so one environment can plug into different trainers and harnesses. It's run by a multi company technical committee (Hugging Face, Meta PyTorch, Microsoft, NVIDIA and others), and next on the roadmap are environment validation and training recipes (AAIF blog)…

Alibaba hands OpenSandbox, its self-hosted sandbox runtime for AI agents, to the AAIF

Announced Oct 22, the Apache 2.0 project (about 16K GitHub stars) gives coding, GUI and browser agents isolated environments to run commands, edit files and execute code on Docker locally or Kubernetes at scale, with SDKs for Python, Java/Kotlin, TypeScript, .NET and Go plus an MCP server. Its sandbox protocol defines lifecycle and execution APIs so teams can plug in custom runtimes, and the repo now lives at opensandbox group/OpenSandbox…

Spotted something we missed? Start a thread.