OpenAI Developers announced Agent Plugins on August 6, 2026 at roughly 9:41pm PT: "Build a plugin once and use it across compatible agent clients. Introducing Agent Plugins, an open standard developed with AWS, Cursor, GitHub, VS Code, and Vercel — that packages Agent Skills and supports MCP server configurations in a shared format." The post crossed 1.1M+ views within a day, and community reaction to the standard itself was strongly positive.
If you already work with Agent Skills or MCP servers, this matters directly: Agent Plugins is not a third, competing format — it's the packaging layer that sits on top of both, so the skill-plus-tooling bundle you build once can install into six different agent clients without rewriting per-client setup instructions.
This is one day old as of publication. Treat the sections below as what's confirmed from the official announcement — and read the honest "what's still unknown" section before you build anything on top of it.
TL;DR — the questions developers are actually asking
| Question | Direct answer |
|---|---|
| What is an Agent Plugin? | A shared packaging/manifest format that bundles an Agent Skill with MCP server configuration into one distributable unit |
| Is it different from an Agent Skill? | Yes — a skill is instructions; a plugin is the container that packages a skill (plus MCP config) for cross-client distribution |
| Is it different from MCP? | Yes — MCP standardizes live tool connections at runtime; Agent Plugins standardizes how a skill+MCP bundle is packaged and installed |
| Who built it? | OpenAI, AWS, Cursor, GitHub, VS Code (Microsoft), with Vercel as a listed collaborator |
| Where's the spec? | agent-plugins.org, per the OpenAI Developers announcement |
| Which clients support it at launch? | Codex, ChatGPT, Cursor, GitHub Copilot, Kiro, VS Code |
| Do I rewrite my existing skills/MCP servers? | Not implied by the announcement — it reads as a wrapping format, but exact migration steps weren't published on day one |
| Is it a rival to Anthropic's Agent Skills format? | No — OpenAI's own description says Agent Plugins "packages Agent Skills," i.e. it wraps that format rather than replacing it |
| Is full spec documentation available yet? | Not as of Aug 7, 2026 — this is a one-day-old standard; expect the spec to keep filling in |
What is an Agent Plugin, exactly?
An Agent Plugin is a packaging standard, not a new runtime or a new instructional format. Based on OpenAI Developers' own description, it bundles two things that already existed separately in the agent-tooling ecosystem:
- Agent Skills — packaged instruction sets (the SKILL.md-style pattern explainx.ai has covered extensively: progressive disclosure, portable playbooks, rules vs. skills vs. MCP)
- MCP server configurations — the connection details for Model Context Protocol servers that give the agent live tools, resources, and prompts
Agent Plugins packages both into "a shared format" — a manifest a plugin author writes once, and that any compatible agent client can read and install from, instead of each client requiring its own bespoke setup flow for "install this skill and wire up this MCP server."
Think of the layering this way:
| Layer | What it standardizes | Analogy |
|---|---|---|
| MCP | How an agent talks to a live tool/data source at runtime | The USB-C port |
| Agent Skills | How instructions/procedures are packaged for an agent to load | The instruction manual |
| Agent Plugins | How a skill + its MCP config ship and install together, across clients | The retail box the manual and the cable both come in |
Do I need to rewrite my Agent Skills or MCP servers?
The announcement language is reassuring here but not exhaustive. OpenAI Developers describes Agent Plugins as something that "packages Agent Skills and supports MCP server configurations" — phrasing that reads as a wrapper around existing assets, not a mandate to rebuild them in a new format.
What's genuinely unclear one day in: the exact manifest shape (is there a plugin.json or equivalent?), whether there's a CLI to auto-generate a plugin from an existing SKILL.md and MCP config, and what versioning/discovery conventions apply. None of that was in the Aug 6 post. Check agent-plugins.org — the official spec site named in the announcement — before packaging anything for production use, since the spec is still filling in.
Which agent clients support Agent Plugins today?
Six compatible clients were named at launch:
| Client | Maker |
|---|---|
| Codex | OpenAI |
| ChatGPT | OpenAI |
| Cursor | Anysphere |
| GitHub Copilot | GitHub / Microsoft |
| Kiro | AWS |
| VS Code | Microsoft |
That list maps almost exactly to the five co-developing organizations named in the announcement — OpenAI, AWS, Cursor, GitHub, and VS Code — plus Vercel, which is credited as a collaborator on the standard even though no Vercel-branded agent client shipped in the day-one compatibility list. This is a genuine multi-vendor coalition, not an OpenAI-only spec dressed up with partner logos — the same pattern explainx.ai covered when GitHub's Copilot SDK shipped a shared agent runtime across six languages, and comparable to how Claude Code, Cursor, and GitHub Copilot already compete and interoperate on tooling standards.
Why not just extend MCP or Agent Skills instead of a new standard?
Because the two existing standards solve genuinely different, narrower problems, and neither one defines packaging:
- MCP defines the wire protocol and trust model for an agent to discover and call tools at runtime — it says nothing about how a bundle of "this skill plus this MCP server" gets versioned, distributed, and installed as a single artifact.
- Agent Skills defines the instructional package format — but a skill that depends on a specific MCP server has historically needed separate, undocumented setup instructions to wire the two together, repeated per client.
Agent Plugins fills that specific gap: a shared install-time container. The fact that five separate companies co-developed it — rather than OpenAI shipping it alone and hoping others adopt — is the strongest signal that this was shared pain across the ecosystem, not a speculative OpenAI initiative. It also lands the same week as Cloudflare's WebMCP and Agents Week announcements, part of a broader August 2026 pattern of vendors converging on portable, shared standards for agent tooling: MCP for runtime tool-calling, WebMCP for site-exposed tools, and now Agent Plugins for packaging skills and MCP configs together.
Is this a competitor to Anthropic's Agent Skills format?
No. OpenAI Developers' own description says Agent Plugins "packages Agent Skills" — present tense, as a stated capability, not "packages a new alternative to Agent Skills." That framing positions Agent Plugins as a distribution layer built to wrap the existing Agent Skills format, the same one explainx.ai's complete guide to agent skills documents in depth, rather than a rival instructional spec competing for the same job.
What's still unknown (be honest about day-one coverage)
This is genuinely new — the announcement is roughly 24 hours old as of publication. Treat the following as open questions, not settled facts:
- Exact manifest schema — field names, required vs. optional keys, versioning scheme were not detailed in the OpenAI Developers post.
- Packaging tooling — whether there's an official CLI (comparable to
npx skills initin the Agent Skills ecosystem) to generate a plugin manifest wasn't stated. - Discovery/registry — whether a central plugin registry exists or is planned, similar to how MCP has
registry.modelcontextprotocol.io, is unconfirmed. - Security/consent model — MCP has a documented consent model for tool invocation; whether Agent Plugins adds its own permission layer on top, or simply inherits the underlying MCP server's model, isn't specified yet.
- Versioning and updates — how a plugin author ships an update, and how clients detect and apply it, is undocumented in the announcement.
Check agent-plugins.org directly for the current state of the spec before shipping anything that depends on these details — this article reflects what was stated in the Aug 6, 2026 announcement, not assumptions about how the spec will fill in.
Not to be confused with: OpenAI's earlier "Codex plugins" repository
If you've been tracking OpenAI's Codex ecosystem, note that this is not the same thing as the official Codex plugins repository OpenAI shipped back in June 2026 — that was a curated GitHub repo of example Codex plugin integrations (Figma, Notion, iOS, web deployment). Agent Plugins, announced August 6, 2026, is the formal open standard those informal examples now sit under — a shared, multi-vendor manifest format rather than a single-vendor example repo. It's also distinct from the separate Codex plugin for Claude Code, which is a Claude Code marketplace plugin that lets Claude Code call out to the Codex CLI — a different "plugin" concept entirely, scoped to one client rather than a cross-client standard.
Bottom line
Agent Plugins is the packaging layer the agent-tooling ecosystem has been missing: a shared manifest that bundles an Agent Skill with its MCP server configuration so the pair ships and installs as one unit across Codex, ChatGPT, Cursor, GitHub Copilot, Kiro, and VS Code. It doesn't replace MCP or Agent Skills — it wraps both. The multi-vendor coalition behind it (OpenAI, AWS, Cursor, GitHub, Microsoft, Vercel) is the strongest signal this addresses real, shared pain rather than one company's agenda. What's missing one day in is spec depth: manifest schema, tooling, registry, and security model are all still filling in at agent-plugins.org. explainx.ai will update this guide as the spec matures.
Related reading
- What are agent skills? Complete guide for Claude Code, Cursor & MCP
- What is MCP? Model Context Protocol: Complete Architecture Guide
- GitHub Copilot SDK: multi-platform agent development
- Claude Code vs Cursor vs GitHub Copilot
- OpenAI's official Codex plugins repository (Figma, Notion, iOS, web)
- OpenAI Codex plugin for Claude Code — setup guide
- CLAUDE.md vs SKILL.md vs MCP: the modern agent stack
- Browse the explainx.ai skills registry →
- Browse the explainx.ai MCP server directory →
Official source: OpenAI Developers on X, August 6, 2026 · spec site: agent-plugins.org.
This article reflects the Agent Plugins standard as announced on August 6, 2026 — one day before publication. Spec details, tooling, and client support may change; check agent-plugins.org for the current state before building against it.