- Home
- Integrations
- Agent plugins
Claude Code, Cursor, Codex and Gemini CLI
SnapRender plugins for coding agents
Install SnapRender, the screenshot API for AI agents, inside your coding agent. One plugin adds the hosted SnapRender MCP server and three skills, so the agent can see a web page on desktop and a phone, read a web page as markdown before it writes code, and set up scheduled screenshots and change monitoring. You sign in through the browser, with no key to paste.
200 free captures a monthNo cardFree plugin, MIT
/plugin marketplace add User0856/snaprender-agent-plugin
/plugin install snaprender@snaprendercodex plugin marketplace add User0856/snaprender-agent-plugin
codex plugin add snaprender@snaprender
codex mcp login snaprendergemini extensions install https://github.com/User0856/snaprender-gemini-extensionnpx skills add User0856/snaprender-agent-plugin
Cursor and other clients: install steps below.
Install in your agent
Pick your client. Each install adds the SnapRender MCP server and the three skills; the first time the agent uses SnapRender, you sign in or create a free account in the browser.
-
Claude Code
Plugin from the SnapRender plugin marketplace: the MCP server and the three skills.
Then run
/mcp, pick the SnapRender server and sign in, or create a free account, in the browser. The skills show up as/snaprender:see-web-page,/snaprender:read-web-pageand/snaprender:watch-web-page, and Claude also picks them on its own when a request fits.Only want the MCP server? Claude Code MCP setup
Inside Claude Code/plugin marketplace add User0856/snaprender-agent-plugin /plugin install snaprender@snaprenderOr from your shellclaude plugin marketplace add User0856/snaprender-agent-plugin claude plugin install snaprender@snaprender -
Cursor
Plugin with the MCP server, the three skills and the logo. It installs from the plugin folder until it is listed in the Cursor Marketplace.
Copy the repository's contents into that folder, so the manifest ends up at
~/.cursor/plugins/local/snaprender/.cursor-plugin/plugin.json. Restart Cursor or run Developer: Reload Window, open Customize, find the SnapRender MCP server and choose Authenticate, then sign in to SnapRender in the browser and approve.Only want the MCP server? Cursor MCP setup
Plugin folder~/.cursor/plugins/local/snaprender/Or only the MCP server, in your Cursor MCP configuration{ "mcpServers": { "snaprender": { "url": "https://app.snap-render.com/mcp" } } } -
Codex
Plugin from the same marketplace: the MCP server and the three skills.
The login command opens the SnapRender sign-in in your browser.
Only want the MCP server? Codex MCP setup
In your shellcodex plugin marketplace add User0856/snaprender-agent-plugin codex plugin add snaprender@snaprender codex mcp login snaprenderOr only the MCP servercodex mcp add snaprender --url https://app.snap-render.com/mcp codex mcp login snaprender -
Gemini CLI
Extension with the MCP server, the three skills and a GEMINI.md context file that tells Gemini when to use each tool.
Restart Gemini CLI. The first time Gemini uses the SnapRender server it opens your browser to sign in and approve. Inside Gemini CLI,
/mcp auth snaprenderrepeats the sign-in and/mcpshows the server and its tools. Update later withgemini extensions update snaprender.Only want the MCP server? Gemini CLI MCP setup
In your shellgemini extensions install https://github.com/User0856/snaprender-gemini-extension -
Skills only (skills.sh)
The three skills, for any agent that reads Agent Skills: Claude Code, Codex, Cursor, OpenCode and many more.
The CLI asks which skills to install and for which agents. The skills call the SnapRender MCP tools, so connect that agent to the MCP server too.
In your shellnpx skills add User0856/snaprender-agent-plugin
Source and READMEs: snaprender-agent-plugin (Claude Code, Cursor, Codex, skills) and snaprender-gemini-extension on GitHub. Plugin version 1.0.0.
The three skills
Skills tell the agent when to reach for SnapRender and how to use it well. In Claude Code they show up as /snaprender:<skill>, and the agent also picks them on its own when a request fits.
-
See a web page
see-web-page
Screenshots a page on desktop and on a phone, looks at both images and reports layout problems: cut-off text, overlaps, broken navigation, unreadable text, missing images.
Use it for
How a page looks, what breaks on mobile, whether a deployed UI change worked.
When not to use it
- Local or private addresses such as localhost or an intranet: ask for a public or preview URL.
- Pages behind a login: SnapRender has no access to your signed-in session.
- Clicking through a flow such as sign-in or checkout: use a browser automation tool. SnapRender clicks one element at most.
- When only the words matter: read-web-page is lighter on context than an image.
Tools it calls
take_screenshotTryScreenshot https://example.com on desktop and on iPhone 15 Pro and list what looks broken on mobile
-
Read a web page as markdown
read-web-page
Reads a page as markdown, trimmed to the main content, so the agent works from the current docs instead of memory.
Use it for
Before writing code against online docs, an API reference or a changelog, or to summarize or quote a page.
When not to use it
- Local or private addresses: read the files in the workspace instead.
- Pages behind a login: paste or export the text instead.
- Content that appears only after typing or clicking: use a browser automation tool.
- Layout or visual questions: use see-web-page.
Tools it calls
extract_contentTryRead the docs at https://example.com/docs as markdown and summarize the API before we write code
-
Scheduled screenshots and change monitoring
watch-web-page
Sets up scheduled captures that compare each capture with the one before and alert you when the page changes. Schedules are created in the SnapRender dashboard or the REST API; the MCP server creates the webhook for alerts.
Use it for
Watching competitor pricing, terms, a client site, or your own pages after deploys.
When not to use it
- Local or private addresses, or pages behind a login: a schedule cannot reach them either.
- A one-off comparison now: take two screenshots with see-web-page.
- Changes that appear only after several clicks: a schedule captures one page.
- Pages that change on every load, such as live feeds: hide that part or raise the change threshold.
Tools it calls
create_webhook,test_webhook,get_usageTryWatch https://example.com/pricing every day and tell me how I get alerted when it changes
watch-web-page sends you to the scheduled captures dashboard page or the REST API to create the schedule, because the MCP server has no schedule tool. Every plan includes schedules; the free plan runs one daily or weekly schedule.
Any other MCP client
No plugin for your client? Point it at the hosted server. You get the same 11 tools; only the skills are missing.
https://app.snap-render.com/mcp- Clients with OAuth open the SnapRender sign-in in the browser.
- Clients without OAuth send
Authorization: Bearer <your API key>. Create a key on your dashboard. - Local server instead:
npx -y snaprender-mcpwithSNAPRENDER_API_KEYset to your key.
| Tool | What it does |
|---|---|
take_screenshot | A page as PNG, JPEG, WebP or PDF, on desktop or a device preset, one screen or the full page |
extract_content | A page as markdown, text, HTML, article, links or metadata |
batch_screenshots | Up to 50 URLs in one background job |
get_batch_status | A batch job's progress and download links |
check_screenshot_cache | Whether a free cached copy exists, without capturing |
get_usage | Captures used and left this month, and the reset date |
sign_screenshot_url | A signed link that renders a screenshot when opened |
list_webhooks, create_webhook, delete_webhook, test_webhook | Webhook notifications, including change alerts from scheduled captures |
Good to know
What the agent can reach, and what the plugin does on your machine.
Public pages only
SnapRender cannot reach localhost, private networks or pages behind a login, and it clicks one element at most before a capture. Give the agent a public or preview deployment URL.Firewall checks
If a capture of your own site shows a security check such as "Just a moment...", one rule lets SnapRender through: allow SnapRender.Nothing runs locally
The plugin adds the MCP server and three instruction files. Only the URL and capture options the agent passes are sent; SnapRender never reads your code or files.Ads and cookie banners
Removed from captures by default, so the agent sees the page and not the consent dialog. Setblock_cookie_bannersto false to keep them.Free checks
get_usageandcheck_screenshot_cachecost nothing, and failed captures are not counted.Captures as files
For apps that save files from tool results,take_screenshottakesas_file: trueand returns the capture as an attached file.
Sign in once, from the agent
Install the plugin, then sign in or create a free account when the browser opens. No key to copy.
200 captures a month free, no card. Every plan works with the plugins.
SnapRender plans
The plugins are free. Each screenshot or extraction uses one capture from your plan.
| Plan | Price a month | Captures a month |
|---|---|---|
| Free | $0 | 200 |
| Starter | $9 | 2,000 |
| Growth | $29 | 10,000 |
| Business | $79 | 50,000 |
| Scale | $199 | 200,000 |
The free plan has 200 captures a month (25 until you verify your email; Google and GitHub sign-ins start at 200). Yearly billing costs ten months of the monthly price. All plans.
Questions
What is the SnapRender agent plugin?
A free plugin that installs SnapRender, the screenshot API for AI agents, inside Claude Code, Cursor and Codex, with a matching extension for Gemini CLI. It adds the hosted SnapRender MCP server (11 tools) and three skills: see a web page, read a web page as markdown, and set up scheduled screenshots and change monitoring.
Do I need an API key?
Not in Claude Code, Cursor, Codex or Gemini CLI: they open the SnapRender sign-in in your browser and keep the tokens. On a machine without a browser, such as over SSH or in a container, create a key on your dashboard and send it as Authorization: Bearer sk_live_... to https://app.snap-render.com/mcp, or run the local server with npx -y snaprender-mcp and SNAPRENDER_API_KEY set.
What does it cost?
The plugin, the extension and the skills are free and MIT licensed. Captures come from your SnapRender plan: Free $0 for 200, Starter $9 for 2,000, Growth $29 for 10,000, Business $79 for 50,000, Scale $199 for 200,000 captures a month. The free plan has 200 captures a month (25 until you verify your email; Google and GitHub sign-ins start at 200), no card. Each screenshot or extraction uses one capture; failed renders, get_usage and the cache check cost nothing.
Is it in the Cursor Marketplace?
Not yet. Until it is listed, copy the repository into ~/.cursor/plugins/local/snaprender/ and reload Cursor, or add only the MCP server to your Cursor MCP configuration. Both are on this page.
Can my agent screenshot localhost or my staging site?
SnapRender reaches public pages only, so localhost, private networks and pages behind a login are out of reach. Use a public or preview deployment URL. If your own site shows a security check instead of the page, one firewall rule lets SnapRender through: see allow SnapRender on your site.
What does the plugin send and store?
The plugin runs no code on your machine. It adds the MCP server and three instruction files. When the agent calls a SnapRender tool, the URL and capture options it passes are sent to app.snap-render.com; SnapRender does not read your code, files or workspace. Captures are cached for a limited time so repeat requests can be served free. Details in the privacy policy.
Can I use the skills without the plugin?
Yes. npx skills add User0856/snaprender-agent-plugin installs them into the agents you pick, including OpenCode and others that read Agent Skills. The skills call the SnapRender MCP tools, so connect that agent to the MCP server too: setup for each client.