OpenAI has added a plugin system to Codex, allowing direct integration with tools like Slack and Figma.

SnapshotBot · 2026-03-28T11:25:00+00:00

OpenAI has launched the Codex plugin system, supporting integration with tools like Slack, Figma, Notion, and others, aiming to simplify the integration process of AI agents. The plugins are designed to be portable and reusable, reducing development costs. Compared to Claude Code, Codex emphasizes official control and security, but third-party compatibility remains to be seen. It is suitable for early adopters and teams with technical expertise.

SnapshotBot

2026-03-28 11:25:00

Abstract generation in progress

Headline

OpenAI has added a plugin system to Codex, now allowing direct connections to commonly used development collaboration tools like Slack, Figma, and Notion.

Summary

OpenAI has launched the Codex plugin, packaging skills, application integrations, and the MCP server into portable modules that can be used in Codex App, command-line, and IDE extensions. It currently supports Slack, Figma, Notion, Gmail, Google Drive, and also provides @plugin-creator for developers to create their own plugins. According to the official documentation dated March 27, 2026, this means Codex can now perform tasks beyond code generation—such as summarizing emails and extracting design specifications. Teams no longer need to write integration code for each tool separately, saving a lot of repetitive work.

In-Depth Analysis

The development path of Codex is quite clear: starting from code completion, to GPT-5-Codex incorporating automation capabilities, and now moving towards multi-step workflows. The plugin mechanism pushes its capability boundaries further, allowing stable connections to external applications, and the connection capabilities can be reused.

The problem it aims to solve is that previously, integrating AI agents into actual work environments required separate integration development for each project and tool, which was quite cumbersome. The plugin abstracts and packages this connection, allowing different teams and projects to share and reuse it, thus reducing marginal costs. OpenAI describes it in the developer documentation as “a reusable packaged unit that can be shared across projects.”

Competition with Claude Code has become more direct: Claude Code already supports a MCP-based integrated ecosystem; OpenAI’s chosen path is “an officially curated plugin marketplace + one-sided control.” This design may be more appealing to enterprises in terms of security, compliance, and maintainability, but it raises a question: how much space will there be for third-party and open-source ecosystems within this framework? It remains unclear.

A simple comparison:

Dimension	Codex Plugin	Claude Code (MCP Integration)
Integration Path	Officially curated plugin marketplace, one-sided control	Native support for MCP integration
Reuse Method	Officially described as “reusable packages that can be shared across projects”	Integration focused on the MCP ecosystem
Ease of Use	Teams may find rapid integration and governance easier	Depends on MCP capability and ecosystem maturity
Openness	Compatibility with third-party and open-source tools still needs observation	Following an open protocol route (MCP)

Impact Assessment

Importance: High
Category: Product Release / Developer Tools / Market Impact

Judgment: The AI transformation of multi-tool workflows is still in its early stages, suitable for teams and ecosystem builders willing to take the first step. Development teams with internal engineering accumulation will benefit first; if you are a transactional participant or long-term capital, it is better to wait and observe until the ecosystem and openness boundaries are clearer.

View Original

This page may contain third-party content, which is provided for information purposes only (not representations/warranties) and should not be considered as an endorsement of its views by Gate, nor as financial or professional advice. See Disclaimer for details.