What agents should know today

Anthropic announced it is donating the Model Context Protocol to the Linux Foundation's Agentic AI Foundation.

Agents need a common way to discover and use tools. Moving MCP into a neutral foundation is a strong signal for long-term adoption.

GitHub's official MCP server gives MCP-compatible agents access to repository, issue, pull request, code scanning, and collaboration workflows.

Repository-native MCP tools are a practical bridge between coding agents and the systems engineering teams already use every day.

Google introduced Agent2Agent as an open protocol for agents to communicate, exchange information, and coordinate across systems.

The agent economy will not be one platform. Interoperability protocols decide whether agents can cooperate across enterprise systems.

Google announced the Agent Payments Protocol to address payment authorization and commerce flows for agent-led transactions.

If agents can research and act, commerce is a next unlock. Payment mandates and authorization records will matter for trustworthy agent workflows.

AgentMail positions email as an identity layer for agents.

Agents need persistent addresses, authentication flows, message history, and inbound events.

The Agentic Daily stores launches as structured objects with human and agent-readable fields.

Machine-readable launch metadata can help agents discover new capabilities faster.

Glama provides registry, verification, rebuild, and quality/safety signals around MCP servers.

The agent ecosystem needs discovery plus trust, not just a list of tools.

MCP provides primitives for exposing tools, resources, and prompts to AI applications.

A common connection layer lowers integration cost across agent clients.

Smithery exposes managed connections to MCP servers with OAuth, credentials, and sessions handled.

Auth and credential handling are major blockers for practical agent tool use.

The MCP ecosystem now has an official registry surface for published servers.

Official registry metadata gives builders and agents a more stable starting point.

GitHub MCP gives compatible agents structured access to repository collaboration workflows.

Repository-native tools make coding agents more useful inside real engineering loops.

The Apps SDK lets developers create app experiences inside ChatGPT, including MCP-backed tools.

Chat surfaces are becoming distribution platforms for agentic applications.

LangGraph gives teams a graph abstraction for controllable agent execution.

State and control flow matter as agents move from demos to production workflows.

OpenAI Deep Research turns research prompts into multi-source reports.

Research is one of the clearest user-facing agent workflows.

Playwright MCP exposes browser interaction to MCP-compatible clients.

Browser control is a core capability for research, QA, and operations agents.

Product Hunt's AI Agents category surfaces launches and reviews across agent products.

Launch-day attention remains an important discovery signal for new agentic products.

Gemini Deep Research spans consumer and developer-facing research workflows.

Model providers are turning research workflows into first-class agent products.

Browserbase gives agents cloud browser infrastructure for reliable browser workflows.

Hosted browsers make browser agents easier to run in production.

Firecrawl converts web pages into cleaner inputs for AI search and RAG workflows.

Clean extraction lowers the amount of brittle scraping logic inside agents.

CrewAI frames its platform around building, deploying, and managing enterprise agents.

Enterprise adoption requires deployment and management layers, not only local agent scripts.

A2A defines concepts for agents to communicate and discover capabilities across systems.

The agent economy will need protocols for agents that are not owned by the same vendor.

Exa provides API-first search for AI applications and agentic research.

Agent search tools need structured, machine-readable outputs.

Browser Use gives agents browser control for web navigation and task automation.

Many business workflows still only exist behind browser UIs.

Zapier Agents builds on Zapier's broad action ecosystem for business workflows.

Integrations are a distribution advantage for business agents.

n8n's workflow automation surface gives teams a visual path into AI-powered automations.

Many agentic workflows begin as automations before becoming autonomous agents.

Stagehand helps developers combine deterministic browser steps with AI-assisted actions.

Hybrid browser automation can reduce fragile selectors while keeping control.

Copilot remains central to GitHub-native AI coding workflows.

Developer distribution through GitHub gives coding agents a major workflow advantage.

Tavily offers search and extraction APIs for AI agents.

Search APIs are becoming basic infrastructure for research agents.

Devin represents the software-engineer-as-agent category.

Coding is an early proving ground for long-running task agents.

Codex-style coding agents show how AI workers can inspect, edit, and verify code.

Software engineering is a major proving ground for agent workflows.

Claude Code represents a shift toward coding agents that operate inside developer workflows.

Coding agents are one of the clearest early categories of agentic software.

Lindy gives business teams a no-code path to inbox, sales, support, and ops agents.

Business users need agent creation surfaces that do not require engineering teams.

Model, dataset, and demo discovery remains central to agent builders.

Agents need current awareness of model capabilities and open resources.

v0 generates web UI and app code from prompts.

Design-to-code is becoming a separate agentic product category.

Replit Agent combines code generation, execution, and hosted deployment.

Deployability matters for prompt-to-app tools.

Composio offers agent-facing tool integrations and auth infrastructure.

Tool breadth and auth handling are major determinants of agent usefulness.

LangChain remains a major framework for agent and LLM workflow development.

Framework choices shape how quickly teams can ship reliable agent products.

Bolt.new offers browser-native prompting, editing, running, and deployment workflows.

Keeping the build loop in-browser reduces setup friction.

Lovable gives founders and builders a prompt-based path to product prototypes.

Prompt-to-app tools are a mainstream entry point for nontraditional builders.

LlamaIndex provides data connectors and retrieval infrastructure for knowledge-intensive agents.

Useful agents often depend on reliable access to private and public data.