Percept
Agent observability platform that discovers and recommends MCP servers, with GA4, Vercel, GitHub and Sentry connectors and ROI tracking.
Install
The installation steps are in the documentation below. Documentation
Documentation
From the project's README on GitHub, refreshed weekly. Links open on GitHub.
perceptdot
AI Visual QA. visual_check(url) screenshots a page with Cloudflare Browser Rendering and finds layout bugs with Gemini 2.5 Flash. Use it from any MCP client (Claude Code, Cursor, Windsurf, Cline) or as a GitHub Action.
A typical page takes about 7 seconds, long pages up to ~20 seconds. Fully headless: no desktop or GUI needed.
Quick start
- Get a free API key at perceptdot.com (100 credits per month, no card).
- Add the remote MCP server. The key goes in the
X-Percept-Keyheader.
Claude Code
claude mcp add --transport http perceptdot https://mcp.perceptdot.com/mcp --header "X-Percept-Key: YOUR_KEY"
Cursor (~/.cursor/mcp.json)
{"mcpServers":{"perceptdot":{"url":"https://mcp.perceptdot.com/mcp","headers":{"X-Percept-Key":"YOUR_KEY"}}}}
Windsurf (~/.codeium/windsurf/mcp_config.json)
{"mcpServers":{"perceptdot":{"serverUrl":"https://mcp.perceptdot.com/mcp","headers":{"X-Percept-Key":"YOUR_KEY"}}}}
Then ask your AI: "Check https://your-site.com with perceptdot".
GitHub Action
The action lives in the eye-action/ folder of this repo.
name: Visual QA
on: [pull_request]
jobs:
visual-check:
runs-on: ubuntu-latest
steps:
- uses: perceptdot/percept/eye-action@v1.0.0
with:
url: ${{ steps.deploy.outputs.preview_url }}
api_key: ${{ secrets.PERCEPT_API_KEY }}
# fail_on_issue: true # optional
Inputs: url (required), api_key (required), fail_on_issue (optional).
What it checks
- DOM audit (deterministic): viewport overflow, broken images, parent overflow, low contrast.
- AI review: each tile of the page is reviewed by Gemini 2.5 Flash for layout problems, with severity (high / medium / low).
Desktop (1280px), tablet (768px) and mobile (375px) viewports are supported.
Pricing
1 credit = 1 tile = a 1280×1600 px slice of the page. A short page is about 1 credit, a long landing page about 3–5 credits.
| Plan | Price | Credits / month | Status |
|---|---|---|---|
| Free | $0 | 100 (+100 one-time bonus after submitting feedback; resets on the 1st, UTC) | Available |
| Pro | $19/month | 10,000 | Coming Soon |
| Team | $49/month | 30,000 (fair use) | Coming Soon |
Payments are handled by Paddle.
Privacy
Screenshots are not stored. They are sent to Google (Gemini API) and Cloudflare (Workers AI, fallback) for analysis. Analysis results (summary and issues, never screenshots) are cached for up to 10 minutes per URL and viewport. Google may use data sent to unpaid Gemini API tiers to improve its products, so do not check pages that contain personal data.
API
- OpenAPI spec: perceptdot.com/openapi.json
- Docs: https://perceptdot.gitbook.io/perceptdot
- Endpoints:
POST /v1/eye/check,GET /v1/quota,POST /v1/free-keyathttps://api.perceptdot.com
curl -X POST https://api.perceptdot.com/v1/eye/check \
-H "Content-Type: application/json" \
-H "X-Percept-Key: YOUR_KEY" \
-d '{"url":"https://example.com"}'
Legacy packages
@perceptdot/ga4, @perceptdot/vercel, @perceptdot/github, @perceptdot/sentry and @perceptdot/core belong to an earlier product and are no longer maintained.
License
MIT. See LICENSE.
Frequently asked questions
What is Percept?
Agent observability platform that discovers and recommends MCP servers, with GA4, Vercel, GitHub and Sentry connectors and ROI tracking.
How do I install Percept?
Follow the installation steps in its documentation, shown on this page and in its repository.
Is Percept open source?
Yes. Its code is on GitHub under the MIT license.
Who maintains Percept?
It is published by perceptdot. AI.tn lists it but does not maintain or endorse it.
From the community awesome-mcp-servers lists and the project's GitHub repository. Listing is not an endorsement: review a server's code and permissions before giving it access to your data.