Cursor Plugins: The Official Plugin Spec and Catalog for the Cursor AI Editor

What cursor/plugins is and who it is for

cursor/plugins is the official home of Cursor's plugin specification and the growing catalog of plugins built to that specification. The README describes the repository as holding official Cursor plugins for popular developer tools, frameworks, and SaaS products. Each plugin is a self-contained directory at the repository root with its own .cursor-plugin/plugin.json manifest. The target audience is developers who use the Cursor AI editor and want to extend it with structured agent behaviors for tools they already use, or who want to build new plugins following the official format. The repository is organized into first-party plugins written by Cursor, plugins contributed by named individuals, and a third_party/ subdirectory containing integrations with external services.

The plugin manifest architecture: .cursor-plugin/plugin.json

The key structural decision in the plugin format is that each plugin is a directory at the repository root containing a .cursor-plugin/plugin.json manifest. The manifest carries the plugin name, description, author, and category, as reflected in the README's catalog table. Plugins are self-contained, meaning they do not require a shared runtime or a central service call to activate; the manifest and any supporting files live together in the directory. The schemas/ directory at the repository root presumably holds JSON Schema definitions for the manifest format, and the scripts/ directory holds tooling for the repository itself. The create-plugin plugin, written by Cursor, is described as scaffolding and validating new agent plugins, which suggests a programmatic path for creating conformant plugins without writing the manifest from scratch.

Official developer tool plugins: a concrete tour

Several of the Cursor-authored plugins address specific gaps in agent workflows. Thermos performs deep security and correctness audits using harsh code-quality rubrics and parallel subagents, and optionally produces merge-ready PR flows. Orchestrate fans large tasks out across parallel cloud agents using planners, workers, and verifiers with structured handoffs. Advisor consults a stronger model before major decisions, when stuck, and before declaring done. Ralph Loop implements iterative self-referential AI loops using what the README calls the Ralph Wiggum technique. Agent Compatibility runs CLI-backed repository compatibility scans and audits startup, validation, and documentation against reality. PR Review Canvas renders pull request diffs as review canvases grouped by importance. Together these plugins address patterns that teams encounter when running agents on real codebases: parallel execution, quality gates, and model consultation on uncertain decisions.

Community-authored plugins and the third_party integrations catalog

Several plugins in the repository were written by named contributors rather than Cursor itself. Eric Zakariasson contributed Continual Learning (incremental AGENTS.md memory updates from transcript signals), Cursor Team Kit (CI, code review, and shipping workflows), CLI for Agents (patterns for designing CLIs that agents can run reliably), and Grok Voice (realtime speech-to-speech and speech-to-text integration). Lauren Tan contributed pstack, described as a workflow for writing less but higher-quality code through rigorous, parallelizable agent workflows. The third_party/ subdirectory contains integrations with a wide range of external services. The Google Workspace suite covers Gmail, Drive, Calendar, Docs, Sheets, and Slides. CRM integrations include Salesforce, HubSpot, Intercom, and Clay. DevOps integrations include GitHub and Playwright. Marketing and email tools include Klaviyo, Customer.io, MailerLite, and Brevo. Analytics tools include Semrush and Ahrefs.

How cursor/plugins differs from MCP servers

Model Context Protocol (MCP) servers are another way to extend AI coding agents with external tool integrations. An MCP server runs as a separate process and exposes tools through a protocol layer that agents call at runtime. Cursor plugins take a different approach: the manifest and supporting files are bundled directly in the directory, making the plugin self-contained and repository-portable. There is no separate server process to deploy or maintain. The plugin travels with the repository that references it. The practical implication is that a team can commit a plugin to their own repository alongside application code and have the same agent behavior available to everyone who clones that repository. The trade-off is that complex plugins requiring persistent state or external API calls need to encode that logic within the plugin's instructions and tool definitions rather than delegating to a running process.

Limitations: license status, specification ownership, and missing install steps

The repository does not specify a license. The README includes a plugin catalog table but no installation instructions, no CONTRIBUTING guide visible in the top-level entries, and no versioning scheme for the plugin specification. The absence of a license is a real constraint: anyone building a product that packages or distributes these plugins cannot determine what is permitted without contacting Cursor directly. The plugin specification is defined by Cursor, a private company, which means changes to the manifest format, the schema, or the expected semantics are not subject to community governance. A plugin written today may require updates if Cursor revises the specification. The last push to the repository was on 2026-09-25, reflecting active development, and no GitHub releases have been published.

Plugins for Claude Code users: what the catalog offers

Several search terms associated with this repository reference Claude Code plugins specifically. The Cursor plugin specification is designed for the Cursor editor's agent runtime, but the patterns documented in plugins like CLI for Agents, Orchestrate, and Continual Learning describe general principles for agent workflows that apply beyond any specific editor. CLI for Agents documents patterns for flags, help with examples, pipelines, errors, idempotency, and dry-run modes, all of which are relevant to anyone writing command-line tools that an AI agent will run. Orchestrate documents a structured handoff protocol between planner, worker, and verifier agents that represents a general pattern for multi-agent task decomposition. These plugins are worth examining as reference implementations even if Cursor is not the editor in use.

Editorial conclusion

cursor/plugins is the right starting point for any Cursor user who wants to extend their agent workflow with official integrations for external services or add structured patterns for code review, security auditing, and parallel execution. The license is not specified in the repository, which is a concrete detail to resolve before building on this specification in a commercial product. The plugin manifest format is defined by Cursor as a proprietary tool, so the specification may change without notice.

Frequently asked questions

Official sources

1. cursor/plugins on GitHub
2. Issues
3. README