Skip to content
Dashboard

6 Cursor plugins for building and shipping applications

Content Engineer

Cursor plugins give your coding agent task-specific instructions and connections to the services behind your application. Vercel, GitHub, Sentry, Linear, Figma, and Hugging Face cover different parts of development, from implementing a design to investigating a failed release.

Start with the information your next task needs. Reading local code might be enough to change a component, but checking that change against a design or production error requires context from another service. Choose the plugin that supplies that context and the instructions for using it.

Copy link to headingUnderstand plugins, skills, and MCP connections

Plugins bundle capabilities for Cursor's agent. Some provide instructions, some configure external tools through the Model Context Protocol, and others combine both. These components serve different purposes:

Component

Role in a task

Example

Skill

Guides the agent through a repeatable workflow

Checking an app's deployment configuration

MCP connection

Makes a service's tools and data available

Reading a Sentry event or Linear issue

Hook

Runs configured code at a supported lifecycle event

Loading project context when a session starts

Plugin

Packages components for installation and sharing

Distributing deployment guidance with a service connection

Cursor supports its own plugin format and the portable Agent Plugins format. Compatible clients can share skills and MCP configuration. Cursor-specific rules, commands, and hooks still need client-specific support.

Editor extensions have a separate role. Syntax highlighting and editor panels extend the development environment; the plugins here supply capabilities the agent can use while working on your application.

Copy link to headingInstall and manage Cursor plugins

Use Cursor's Customize page to find a plugin, select Install, and choose project or user scope. Project scope keeps the installation associated with the current project; user scope makes it available across your projects.

Marketplace listings also provide chat commands. For Vercel, enter this in Cursor's agent chat:

/add-plugin vercel

Review the package and complete its connection setup. Installation can supply an MCP configuration, but access to private resources still depends on authentication. Verify that the agent can read a specific resource, such as the intended Vercel project, before requesting changes.

Return to Customize to inspect installed plugins and their components. You can toggle MCP servers there; disabled servers no longer appear in chat. If a plugin has instructions but no working service connection, check its MCP setup before reinstalling it.

Copy link to headingChoose plugins for the work you need to do

The selection includes Vercel's own plugin alongside five packages for services developers use around an application. Each has an installable Cursor package, including GitHub and Linear, whose packages primarily configure MCP access.

Plugin

Useful task

Setup or access requirement

Vercel

Prepare deployments and apply platform guidance

Node.js, Bun, and Vercel authentication

GitHub

Review pull requests and investigate Actions failures

Personal access token with relevant repository permissions

Sentry

Add monitoring and diagnose recorded errors

Sentry sign-in and useful event data

Linear

Implement issues using their acceptance criteria

Access to the correct Linear workspace

Figma

Build UI from frames and existing components

File access and applicable seat limits

Hugging Face

Research models and work with Hub resources

CLI setup for CLI workflows; account access for private resources

Copy link to heading1. Vercel for deployment and platform guidance

The Vercel Plugin combines platform skills with a Vercel MCP connection. It can guide AI SDK work, environment configuration, and deployment tasks. Deployment and performance specialists provide more focused instructions when a task needs them.

Have Node.js 18 or later and Bun available for the plugin's local runtime. After installing it with the chat command above, complete Vercel authentication for project data and deployment tools.

Give Cursor a project-specific request:

Prepare this branch for a Vercel preview. Check the build settings and required environment variable names, then show any changes needed before deployment.

The plugin's default session-start context activates in empty directories and detected Vercel, Next.js, or eve projects. Skills remain available on demand; the default hooks don't inject a skill after every edit. This makes explicit requests useful when you want Cursor to review a particular platform concern.

Before creating the preview, confirm the destination team and project. The plugin's deployment guidance and the connected account's deployment permissions are separate parts of that workflow.

Copy link to heading2. GitHub for pull requests and repository context

Cursor's GitHub plugin connects the agent to repository, issue, pull request, and Actions tools. It is useful when the local checkout lacks the review discussion or CI results that explain a change.

Install it from Cursor's agent chat:

/add-plugin github

The package manifest requires Cursor 3.13 or later and a GitHub personal access token value. For a team-managed installation, an admin supplies that value under Plugins → Configure in the dashboard. Cursor substitutes this plugin variable into the connection configuration.

Limit the token's repository access and permissions to the work you intend to perform. Keep its value in the plugin's configuration rather than committing it to the package repository.

For a failing pull request, try:

Read this pull request's discussion and failed Actions run. Compare the failure with the local branch, reproduce it where possible, and prepare a patch with the relevant test result.

Permission to read a repository doesn't necessarily include permission to post reviews or change it. Check the available operations before asking Cursor to publish a review or update the pull request.

Copy link to heading3. Sentry for instrumentation and production errors

Sentry's plugin pairs its hosted MCP connection with skills for instrumentation and debugging. It can help add monitoring to a new application or work backward from an error to the code that produced it.

Add the package in Cursor chat:

/add-plugin sentry

Complete the hosted server's OAuth connection with the Sentry account that can access your project. Self-hosted Sentry uses a separate CLI-backed connection, so the hosted plugin setup alone doesn't establish that access.

Use a recorded issue to focus the investigation:

Investigate this checkout error. Use its stack trace and breadcrumbs to identify the failing request path, then add a regression test and propose a fix without resolving the Sentry issue yet.

Sentry's skills also cover repairing unreadable stack traces. If an event contains minified frames, fixing source-map uploads may be the necessary first step. The plugin can guide that work, but it cannot reconstruct context that the application never recorded.

Copy link to heading4. Linear for requirements and implementation tracking

The Linear plugin gives Cursor tools to work with issues and projects. Its package configures the service connection, so it can be useful even without a separate library of workflow skills.

Enter the installation command in Cursor chat, then authenticate to the intended workspace:

/add-plugin linear

Use the issue discussion to constrain the implementation:

Read this issue and its comments. Turn the acceptance criteria into implementation checks, compare them with the current branch, and flag decisions that still need a product answer.

This is particularly useful when a ticket has changed since work began. Ask Cursor to identify conflicting requirements before it changes code, then use the completed checks to draft an issue update.

Linear's MCP access options include read-only connections. If your team wants requirements research without issue edits, configure that restriction at the connection level. Asking the agent to avoid writes expresses intent; the connection determines which actions are available.

Copy link to heading5. Figma for design implementation

Figma's Cursor plugin supplies design-to-code and Code Connect skills with its MCP server configuration. It gives the agent structured design context, including components and variables, that a screenshot alone doesn't provide.

Install it in Cursor chat:

/add-plugin figma

Share a frame link and identify the existing code it should use:

Build this account-settings frame using our form components. Preserve the design tokens, identify missing error and loading states, and compare the implementation at narrow and wide viewport widths.

Code Connect mappings can help the agent reuse your implemented components. Without those mappings, name the component directory and conventions in the request so the design doesn't become a second, inconsistent UI library.

The connection only reaches files your Figma account can access. Read limits vary by plan and seat, while image generation and some Figma Make actions use AI credits. Check those requirements if the task includes canvas work as well as reading a design.

Copy link to heading6. Hugging Face for model research and Hub workflows

The Hugging Face plugin brings Hub tooling and AI/ML instructions into Cursor. Its install identifier differs from the marketplace URL. Use this command in agent chat:

/add-plugin huggingface-skills

Start with a bounded research task:

Find candidate reranking models for our documentation search. Compare their licenses and supported languages, then propose an evaluation using our queries before downloading weights or starting compute.

For workflows that use local commands, install and authenticate the Hugging Face CLI. The upstream package uses the hf-cli skill as an entry point to Hub operations; additional workflow skills can be installed as needed. Inspect your installed package before assuming a particular training or evaluation skill is present.

Hub access extends beyond search to repositories and remote workloads. Private resources need the appropriate account access, and Jobs have separate compute charges. Keep model selection separate from permission to launch an evaluation or training run.

Copy link to headingCombine plugins around a verifiable result

For a new account-settings page, use Linear to establish the acceptance criteria and Figma to supply the design. Have Cursor map both to the repository's existing components before changing files. That comparison can expose missing states, such as validation errors or an empty profile, while the implementation is still small.

After the repository's checks pass, use Vercel guidance to prepare a preview. Review the page against the design and exercise the issue's acceptance criteria there. GitHub can supply the pull request context for a review summary that explains what changed and what was checked.

Production debugging has a different starting point. Give Cursor a Sentry event and the relevant release or pull request, then ask for a reproducible failure and a regression test. Resolve the issue only after verifying the fix in the intended environment.

Specify which external changes belong in the task. Preparing a code patch, posting an issue update, and deploying to production are different actions, even when one agent has tools for all of them.

Copy link to headingConnect Cursor to Vercel AI Gateway

Cursor can send supported model requests through Vercel AI Gateway. The Vercel Plugin also includes an ai-gateway skill for building Gateway into your application, such as configuring provider routing and fallbacks. Installing that skill doesn't change Cursor's model provider.

Start provider setup from your terminal with your package manager:

pnpm dlx vercel@latest ai-gateway setup --agent cursor

The command provisions a key and prints the remaining steps. Add --dry-run to preview the setup first. Finish the connection in Cursor Settings → Models:

  1. Put the AI Gateway key in OpenAI API Key.

  2. Enable Override OpenAI Base URL and use the endpoint below.

  3. Use Add model to enter an ID from the Gateway model catalog.

Cursor requires its dedicated compatibility endpoint:

https://ai-gateway.vercel.sh/cursor/v1

Select the model, send a request, and check that it appears in AI Gateway. Tab completions bypass custom keys, and some Agent or Auto modes can bypass the override. The override also makes Cursor's built-in non-OpenAI models unavailable until you turn it off. Requests still pass through Cursor's backend.

Gateway model usage follows AI Gateway billing. Cursor's own-key billing rules can also apply a Cursor Token Rate to eligible requests on Teams and Enterprise plans. Verify usage in both services for your team's chosen configuration.

Copy link to headingShare a plugin with your team

Package a workflow once it has useful instructions to repeat across projects. For example, a preview-review skill could compare a Linear issue's requirements with a deployed branch and produce a review checklist with links to the relevant evidence.

For portable skills and MCP configuration, follow the Agent Plugins authoring guide and use a root manifest. Cursor's native format uses a different manifest location when you need Cursor-specific components:

Package format

Manifest location

Appropriate use

Agent Plugins

plugin.json

Portable skills and MCP configuration

Cursor Plugins

.cursor-plugin/plugin.json

Cursor rules, agents, commands, and hooks alongside skills or MCP

Team marketplaces are available on Teams and Enterprise plans. Admins can import a repository through Dashboard → Plugins & MCPs and choose whether a plugin is optional, installed by default, or required. Share the package's instructions and configuration; each service connection still needs suitable credentials.

Copy link to headingFrequently asked questions

Copy link to headingAre Cursor plugins the same as editor extensions?

No. Cursor plugins package capabilities for the agent, such as task instructions and service connections. Editor extensions can add language support, panels, or other editor features. Choose the package type that matches the part of your workflow you want to extend.

Copy link to headingCan a Cursor plugin contain only an MCP connection?

Yes. Plugin packages can configure an MCP server without adding workflow skills. GitHub and Linear are examples in this selection. You still need to authenticate the connection and grant access to the relevant resources.

Copy link to headingWhy does a plugin install but fail to read my project?

Check the plugin's MCP connection in Customize and confirm that it is enabled. Then check the account or token used by that connection, including access to the specific project or repository. Installing instructions and configuring account access are separate steps.

Copy link to headingDoes the Vercel Plugin route Cursor through AI Gateway?

No. The plugin supplies AI Gateway guidance for development tasks, but Cursor's provider connection requires separate setup. Complete the model settings configuration and verify a request in Gateway before relying on it for usage tracking.

Copy link to headingCan my team require a Cursor plugin?

Team marketplace admins can mark a plugin as required, which installs it automatically and prevents members from uninstalling it. Team marketplaces require a Teams or Enterprise plan. Connected services continue to enforce their own account permissions.

More Build with AI articles

Ready to deploy?