On March 26, 2026, OpenAI launched a plugin marketplace for Codex, its AI coding agent. The launch added more than 20 integrations connecting Codex to tools like Slack, Figma, Notion, Gmail, Google Drive, Linear, Sentry, Box, and Hugging Face.

On the surface, this is a developer story. But the implications reach well beyond software teams.

What the Plugins Actually Do

Codex plugins are installable bundles that package three things together: predefined prompt workflows that shape agent behaviour, connectors to external services, and MCP server configurations that link Codex to remote tools or shared data.

The result is a Codex agent that does not just write code — it can read design files from Figma, post updates to Slack, pull documentation from Notion, and manage tasks in Linear, all within a single automated workflow.

OpenAI drew a deliberate distinction between plugins and skills. Skills are personal or project-specific prompt configurations. Plugins are versioned, reusable, and designed to be shared across teams or published publicly. That versioning detail matters for enterprise adoption — it means teams can standardise their agent tooling in a way that personal prompt libraries cannot.

The Numbers Behind This

Codex reached 1.6 million weekly active users in early March 2026, more than tripling after the launch of GPT-5.3 Codex in February. Weekly token usage grew fivefold over the same period.

Enterprise customers including Cisco, Nvidia, Ramp, Rakuten, and Harvey have deployed the tool across developer teams. OpenAI also announced that self-serve plugin publishing is coming soon, which would open the marketplace to third-party developers building custom integrations.

From Developer Tool to Workflow Platform

The Figma integration is the clearest illustration of where this is heading. Codex can now read from Figma Design files and generate code that implements those designs. It can also turn existing code back into editable Figma designs for iteration. For teams where the designer-to-developer handoff is a persistent time sink, this removes a real step.

The Slack and Notion integrations extend Codex into planning and coordination work that surrounds development — pulling context from discussions, updating documentation, and keeping project records current without manual data entry.

This is the pattern across all major AI platforms right now. The model capabilities are largely commoditised. The differentiator is how deeply the AI can integrate into the workflows a business already runs, and whether teams can share and standardise those integrations at scale.

What This Means for Business

If you have avoided thinking about AI coding tools because your business is not a software company, this shift is worth revisiting.

The line between “AI coding tool” and “AI workflow automation tool” is dissolving. A Codex agent configured with the right plugins can draft a competitive analysis, pull data from multiple sources, format findings into a Notion document, and post a summary to Slack — with no code written by a human. That is not a developer workflow. That is a business operations workflow.

For companies evaluating how to deploy AI agents across their operations, a few things stand out here:

The MCP standard is becoming critical infrastructure. Both the Codex plugins and the SAP-Reltio deal this week rely on the Model Context Protocol to connect agents to external systems. Organisations that understand MCP and build their integration strategy around it are going to move faster than those treating each tool connection as a bespoke project.

Composability is the new competitive edge. The businesses getting the most from AI are not the ones with the fanciest model. They are the ones that have figured out how to connect their tools, data, and workflows into coherent agent pipelines. Plugins, MCP, and workflow standards are the building blocks for that.

Your team needs to understand these tools. Whether you are a developer, a data analyst, or a business operator, the gap between people who can configure and direct AI agents and those who cannot is growing quickly. The practical skills — prompt design, workflow architecture, tool integration — are becoming as important as traditional technical skills.

For a deeper walkthrough of tools like this and how they fit together, the free Working With Claude field guide covers the ecosystem end to end. Get the guide.