MCP

2026-10-26

CODING · 1 SOURCE · InfoQ AI, ML & Data Engineering

Agentic AI Foundation adds four projects (OpenSandbox, Open Responses, Agent Skills, OpenEnv) as MCP discussion shifts to security

At AGNTCon + MCPCon 2026 the Agentic AI Foundation said it expanded from three to ten projects after contributions from Alibaba (OpenSandbox), OpenAI (Open Responses), Anthropic (Agent Skills) and Hugging Face (OpenEnv). The conference emphasized security and identity for agent protocols — including MCP task support in Claude Code, WebMCP origin-trial work from Chrome, and growing attention to attack surface after benchmarks like CyberGym and reported exploit results with GPT-5.6 Sol.

7.0

2026-10-24

CODING · 1 SOURCE · InfoQ AI, ML & Data Engineering

Google makes Data Agent Kit generally available, splits agent guardrails between skills and IAM

Google has made its Data Agent Kit generally available as an open-source, free integration that connects coding agents to over fifteen Google Data Cloud services and ships 37 versioned skills (Apache 2.0). The release emphasizes two control layers—advisory ‘‘skills’’ that teach best practices (e.g., accidental-data-loss-prevention) and enforceable access controls via IAM (including deny rules like blocking mcp.googleapis.com/tools.call)—while noting limits (blocked tools still appear in listings, controls cover only Google MCP servers, and some attributes are gcloud-only).

7.0

2026-10-23

MODELS · 3 SOURCES · TechCrunch AI · The Verge AI · The Decoder

Canva launches 'Craft' image-generation model

Canva released its first in-house image-generation model, Canva Craft, a two-stage system that plans layouts and then generates images up to 2K across ~30 aspect ratios; it was trained on Canva’s design templates rather than photos and aims for accurate object detail and legible text. The model is available to all users via the Dream Lab and through the Canva API, will be supported in third-party chatbots via Canva’s MCP, and Canva says faster, lower-cost variants (Craftfast and Craft-ultrafast) are coming; researchers from recently acquired Leonardo.ai contributed to the work.

7.0

2026-10-19

MODELS · 1 SOURCE · The Decoder

Kling AI releases Kling 4.0 with stereo audio, multilingual lip‑sync and a faster Flash variant

Kling AI launched Kling 4.0, a major update to its AI video generator that supports single‑pass generations up to 30 seconds, 8,000‑token prompts, and a new "Omni Reference" system accepting up to 15 reference assets (images, videos, elements). The release adds two‑channel stereo audio, improved lip sync across more than ten languages, a Creative Recreation template feature, a redesigned interface, MCP and CLI access for automation, and a lighter Kling 4.0 Flash variant for faster, lower‑cost 720p/20s outputs; 4K HDR and longer (up to two minutes) extensions are listed as coming soon, and API pricing/support were not announced.

6.0

2026-10-15

COMPANIES · 1 SOURCE · TechCrunch AI

Figma debuts 'Specs' Markdown editor, code layers public beta, and Payload integration to steer AI agents

Figma introduced Figma Specs, a Markdown-based editor for defining a company’s design language that can reference canvas elements, FigJam diagrams, code layers and libraries, and serve as a central source for in‑product agents. The company also opened its Code Layers feature to public beta, expanded availability of its on‑canvas agent, and integrated the Payload headless CMS into Figma Make, while adding support for over 20 tools via its MCP including Grok Bot and Indian startup Sarvam.

6.0

2026-09-18

RESEARCH · 1 SOURCE · arXiv cs.AI

Position paper proposes virtualizing foundation models with a self‑evolving FMOS

An arXiv position paper (2609.19203v1) argues current AI stacks are fragmented and proposes a Foundation Model Operating System (FMOS) — a self‑evolving system layer that virtualizes FM interactions, orchestrates memory tiers, model selection, resource allocation, verification, and adaptive policy enforcement to give applications the illusion of dedicated, trustworthy FM instances.

7.0

2026-03-16

MODELS · 1 SOURCE · Mistral AI

Leanstral: open-source Lean 4 proof-engineering agent (Leanstral-120B-A6B) released under Apache 2.0

Leanstral, an open-source code agent trained for Lean 4 proof engineering and shipped as Leanstral-120B-A6B, is released with Apache 2.0 weights, an agent mode for Mistral vibe, a free API endpoint, a forthcoming tech report, and a new evaluation suite called FLTEval. The model uses a highly sparse architecture (6B active parameters), supports MCPs via vibe (optimized for lean-lsp-mcp), and is benchmarked on realistic repository tasks where it shows cost and efficiency advantages versus several closed- and open-source competitors.

8.0