New Practice lessons are live in the Playground.See what you can learn

Xpensiv MCP

Give your coding agent a second look.

Your agent built the page. Let Xpensiv check the public preview and bring the findings back into the same conversation, ready for you and your agent to work through.

Before / A public preview

Something feels off.

“Does this look finished?” A preview link and another round of guesswork.

After / A check to work from

Know what to fix next.

Your agent can read the findings, inspect the evidence and work from suggested fixes while you decide what to change.

Workflow illustration. Xpensiv reports findings; you or your coding agent make the changes.

Who it’s for

For the way
you already work.

For builders using Cursor, Claude Code or Codex who want to review a page without moving the fix list between tools.

01

Review a page before launch

Ask your agent to check a public app URL. It uses check_app to start the check, then get_check to read the findings.

02

Work through the fix list

Keep the evidence and suggested fixes in your coding session while you and your agent make the edits.

03

Return to an earlier check

Use list_checks to find previous checks, then get_check to read the status and completed findings.

Get connected

Connect once. Ask for a check.

Create an API key, then add the Streamable HTTP server to your agent. This example connects Claude Code. The console has ready-made configs for Cursor and Codex.

Set XPENSIV_API_KEY to your key first. Keep it private.

claude mcp add --transport http xpensiv https://mcp.xpensiv.si --header "Authorization: Bearer $XPENSIV_API_KEY"

Then ask: Use Xpensiv to check https://your-app.com and show me the findings.

Read the MCP setup guide

Start with your next page

Bring the fix list into your next session.

Pay-as-you-go credits from $20. No subscription. Checks are metered by input tokens; output is free.