One Plugin, Six AI Tools
Agent Plugins 1.0.0 [launched August 6, 2026]
AWS, Cursor, GitHub, Microsoft, OpenAI, and Vercel agreed on something that almost never happens in this industry. They shipped a shared, vendor-neutral format for packaging AI agent extensions, so one plugin can run across all of their tools instead of getting rebuilt five separate times.
What it actually is
A plugin under the new spec is just a folder. Inside it you get a plugin.json manifest, an optional skills/ directory, and an optional mcp.json for wiring up MCP server connections. That last piece is worth noting if you've already spent time learning Model Context Protocol this year. Agent Plugins doesn't replace MCP, it sits on top of it, giving you a standard way to bundle MCP connections together with instructions and skills into one distributable package.
Full spec, JSON schemas, and guides for both plugin authors and client implementers are up at agent-plugins.org.
Why cloud folks should care
At launch, the format already works in six clients: ChatGPT, Codex, Cursor, GitHub Copilot, Kiro, and VS Code. Kiro is the one to notice if you work in AWS. Amazon's own coding agent adopting the same plugin format as Microsoft's Copilot and OpenAI's Codex is a real signal. It means the basic plumbing of "how does an AI tool load an extension" is done being fragmented across vendors.
That matters practically. If your team has built or is planning to build internal AI tooling, custom skills, or agent workflows tied to AWS services, you no longer have to pick one client and hope your engineers all standardize on it. Build to the Agent Plugins spec once, and it should run in whichever of the six supported tools your team actually prefers.
Introducing The First Agentic CRM
Get revenue agents, workflows, and automations across every stage of your motion. Access customer data in real time through Attio's web app, MCP, API, and SDK.
Then Ask Attio anything about your business and get instant answers.
It's the CRM that runs the work behind every win.
The governance detail that makes this credible
Plenty of "open standards" from a group of competitors quietly end up steered by whoever proposed them. This one built in guardrails against that. No single company can hold a majority of Core Maintainer seats, and those seats belong to individuals, not employers. Vercel started the proposal, but AWS, Cursor, GitHub, Microsoft, and OpenAI all had a hand in shaping the final 1.0.0 spec before it shipped. That's a decent sign it survives past the initial announcement instead of quietly becoming one vendor's roadmap with an open source label on it.
What to do about it
If you build or maintain AI tooling, agent skills, or MCP integrations for your team, this is the format to target going forward instead of writing one-off integrations per client. Skim the spec at agent-plugins.org and see how close your existing setup already is, most MCP-based tooling is a short step away.
If you're just using these tools day to day and not building extensions yourself, there's nothing to do yet. Just expect "Agent Plugins compatible" to start showing up as a line item on tool comparisons over the next few months, the same way "MCP support" did last year.
Keep Learning and Building.
