Model Context Protocol

Evidence (8)

• 6The New Stack·6/11/2026framework_update
  Designing an AI-first software delivery flow

  The article argues that teams can delegate a significant share of ticket-to-production work to AI if the software delivery process is redesigned around agents. It outlines a five-phase workflow with human approval gates, supported by a context lake, guardrails, and visibility across planning, development, preview, and deploy steps.

• 7The New Stack·6/10/2026framework_update
  Anthropic’s Claude Code creator says agents now need loops

  Anthropic’s Boris Cherny said he no longer prompts Claude directly and instead designs loops that run the work. The article frames this as “loop engineering,” an orchestration pattern built from scheduled execution, isolated workspaces, verifier agents, and persistent memory for long-running coding agents.

• 8The New Stack·6/10/2026open_source
  Databricks launches OpenSharing for secure AI asset sharing

  Databricks launched OpenSharing, the successor to Delta Sharing, as a standalone Linux Foundation project. The protocol extends zero-copy sharing beyond tables to include agent skills, AI models, and unstructured data, while adding support for Apache Iceberg REST Catalog clients and on-prem storage vendors.

• 5InfoQ·6/10/2026research
  Context Engineering and Memory Management for AI at Scale

  Adi Polak outlines how AI systems can move from stateless prompting to state-aware agents with richer context and persistent memory. She highlights Apache Kafka, Flink, and MCP as building blocks for real-time processing, memory tiering, and tool orchestration while addressing token limits, latency, and cost.

• 7InfoQ·6/10/2026product_launch
  Azure API Management Adds Unified Model API and Expanded Safety Controls

  Azure API Management introduced a Unified Model API that allows clients to use a single request format while APIM translates calls to providers such as Anthropic and Vertex AI. Microsoft also expanded content safety policies to cover MCP tool calls and agent-to-agent payloads, while adding token metrics for reasoning, cached, and audio tokens across providers.

• 4InfoQ·6/8/2026research
  AI-Native Engineering Evolves from Vibe Coding to Autonomous Agents

  Birgitta Böckeler discusses how AI use in software delivery has changed over the past year, from vibe coding to a more mature AI-native engineering approach. The conversation also highlights the shifting tool landscape and the rise of more autonomous agents that can increase both productivity and risk.

• 8The New Stack·6/7/2026product_launch
  Microsoft makes OpenClaw the free base for Scout and enterprise agents

  Microsoft launched Scout, its first always-on work agent, on the open-source OpenClaw runtime and contributed enterprise policy controls back upstream. The move positions OpenClaw as a common base while Microsoft focuses on the identity, governance, and management layers above it.

• 4Hugging Face Blog·6/3/2026product_launch
  MCP Tools Added to Reachy Mini

  Hugging Face says Reachy Mini now supports MCP tools, expanding the robot's ability to connect with external capabilities through the Model Context Protocol. The update positions the small robot as a more flexible platform for agentic and tool-using workflows.