New in week 43

Event-based protocol that streams agent state, messages and tool calls between an agent backend and a user-facing app.

JSON-RPC protocol that lets any code editor run any coding agent, locally over stdio or remotely over HTTP.

Open protocol suite for agent identity, discovery and messaging on the open web, built on W3C DIDs.

Open protocol for agent-initiated payments, built as an extension to A2A and MCP, with signed user mandates.

Open folder format (a SKILL.md plus optional scripts and files) for packaging instructions an agent loads on demand.

Open protocol for agents built on different frameworks and vendors to discover each other and collaborate on tasks.

Open checkout standard from OpenAI and Stripe for buyers and AI agents to complete purchases with businesses.

Plain Markdown file at a repository root that tells coding agents how to build, test and work in the project.

Grid puzzles that are easy for people but hard for AI, testing abstraction and reasoning on new problems.

Interactive benchmark of novel game-like environments that agents must explore and learn without instructions.

Leaderboard that tests how accurately models call functions and tools, from single calls to multi-turn agentic use.

1,266 hard-to-find questions with short answers for testing how persistently agents can search the web.

Benchmark of 450+ real-world questions for general AI assistants, split into three difficulty levels.

Real work tasks from professionals across many occupations, graded by experts against human deliverables.

2,500 expert-written, closed-ended questions across dozens of academic subjects, including multimodal items.

Continuously refreshed coding benchmark built from new contest problems to limit training-data contamination.

Proposal for a Markdown /llms.txt file that gives language models a short, curated map of a website's content.

Official MCP extension that lets tools return interactive HTML interfaces rendered in a sandboxed iframe inside the chat.

Official open catalog and API for publicly available MCP servers, run by the MCP project.

Benchmark and framework that tests models and agents on tasks using real-world MCP servers.

Stress test for realistic MCP use across Notion, GitHub, filesystem, PostgreSQL and Playwright servers.

75 Kaggle competitions turned into offline tasks for testing machine learning engineering agents.

Open JSON-RPC protocol that lets AI applications connect to external tools, data sources and prompts through MCP servers.

Framework-neutral configuration language for describing agents and flows so they can be serialized and run on different runtimes.

OpenAI's toolkit for building ChatGPT plugins from MCP servers, skills and optional in-chat UI, now documented under Plugins.

Set of 108 long-horizon computer-use workflows covering everyday and professional desktop work.

Revised version of OSWorld, a benchmark of open-ended tasks on real desktop operating systems for computer-use agents.

OpenTelemetry attribute names and signal shapes for tracing LLM calls, agents, tool execution and MCP operations.

Harder, long-horizon software engineering tasks from public and proprietary codebases, with a held-out set.

Human-validated set of 500 real GitHub issues from Python repositories for testing coding agents.

Simulated customer-service benchmark where an agent follows domain policies and uses tools while a simulated user also acts.

Benchmark of hard, realistic command-line tasks that agents must complete inside a terminal sandbox.

Open standard for agentic commerce covering discovery, checkout and post-purchase between platforms, agents and businesses.

Self-hosted web environment with realistic sites (shopping, forums, code hosting, maps) for testing web agents.

Open payment standard that uses the HTTP 402 status code so clients, including agents, can pay per request.