Let your coding agent set it up.

Run npx lopebase in a product's folder. It reads your stack, connects your agent to LopeBase's MCP server, and your agent adds the product, connects its sources, and sets up uptime while you keep coding.

Or paste one promptInto Claude Code, Cursor, Codex, or Gemini CLI. It runs the setup and asks you for anything only you can do.
See the prompt

Set up LopeBase (lopebase.com) for this project. Run `npx lopebase` in this folder and follow its steps; I will sign in or create a free account in the browser when it asks. Once you are connected to the LopeBase MCP server, call get_onboarding_status and do each step it marks next: add this product as a property, connect the services this codebase already uses (payments, analytics, Search Console, error tracking), add the LopeBase Adapter to the app, and set up uptime checks for the site and the services it depends on. Use request_human_action for anything only I can do, like approving an OAuth screen or pasting a key, and never print secret values back to me.

Every change an agent makes is in the audit log, under the person who created its token.

One command, from your product's folder

npx lopebase does the boring part: it reads your stack, signs this computer in, and hands the rest to your agent with a prompt that says what to do.

$ npx lopebase

Works on macOS, Linux, and Windows with Node 18 or later.

  1. 01

    It scans the folder

    After you say yes, it reads dependency names, which config files exist, and the variable names in files like .env.example. Never your .env file or your code.

  2. 02

    You connect this computer

    A code shows in your terminal and your browser opens. Sign in, or create a free account, check the code matches, and press Connect.

  3. 03

    Your agent takes over

    It finds the agent on your computer, adds the LopeBase tools, and starts it with what the scan found. You watch each step turn green in your workspace.

Using Claude?

LopeBase is listed in Claude's connector directory. Connect in a couple of clicks, no address to copy.

Find LopeBase in the Claude directory

Claude Code by hand

Add the server, then approve the sign-in page it opens.

claude mcp add --transport http lopebase https://app.lopebase.com/api/mcp

Any other MCP client

Add the address as an HTTP server. Clients that cannot sign in take an agent token from Settings, Agent access.

{
  "mcpServers": {
    "lopebase": {
      "url": "https://app.lopebase.com/api/mcp",
      "headers": { "Authorization": "Bearer lb_agent_..." }
    }
  }
}

The help center covers every npx lopebase command and tokens, scopes, and tools.

Works with the agent you already use

npx lopebase finds and starts five coding agents. Anything else that speaks MCP over HTTP connects at one address.

  • Claude CodeStarted by npx lopebase, or added with one command.npx lopebase
  • CodexFound on your computer and started with the setup prompt.npx lopebase
  • Gemini CLIFound on your computer and started with the setup prompt.npx lopebase
  • CursorCursor Agent in the terminal, or the address in Cursor's MCP settings.npx lopebase
  • opencodeFound on your computer and started with the setup prompt.npx lopebase
  • Claude.aiListed in Claude's connector directory.Sign in
  • ChatGPTA custom app in developer mode, on plans that allow custom apps.Sign in
  • Any MCP clientThe address with a sign-in, or an agent token in a header.Address or token

What your agent can read and do

A token carries one or more scopes and only sees the tools they allow. Pick what fits the job.

Setup

Add properties, connect sources, and write drafts.

  • list_propertiesget_propertyYour properties and what each has connected.
  • add_propertyAdd a property by name and site address.
  • connect_sourceConnect an SEO, analytics, revenue, or uptime source, tested live before it is saved.
  • connect_adapterConnect the LopeBase Adapter once it is in your app.
  • detect_stackget_onboarding_statusget_setup_guideWhat a project uses, and what to set up next.
  • get_pageview_snippetget_pageviewsTurn on the cookieless pageview counter, with your yes, and confirm hits arrive.
  • get_content_profilenext_article_topicsubmit_draftWrite an article in a property's voice, through the same checks and review.
  • request_human_actionlist_human_actionsresolve_human_actionHand you a step only you can do, and track it until it is done.

Read

See how things are going. Never changes anything.

  • get_healthWhat needs attention across every property right now.
  • get_digestThis week's numbers against last week.
  • list_metricsquery_metricThe numbers your dashboards show, like MRR, pageviews, and signups.
  • list_sourcesread_sourceSearch Console, Bing, GA4, Plausible, PostHog, and Cloudflare Web Analytics, as totals.
  • get_ai_visibilityHow ChatGPT, Claude, and Gemini answer a property's buyer questions, from the latest check.
  • get_uptime_settingsget_status_pagelist_pipelinesget_discussionsHow uptime, status pages, pipelines, and discussions are set up.
  • get_seolist_articleslist_dashboardsTracked searches, competitors, articles, and dashboards.
  • diagnose_connectionlist_audit_eventsWhy a connection is failing, and what changed and who changed it.

Manage

Change things for you. Grant it only when you want that.

  • set_uptime_settingsHow often each check runs, and when it counts as down.
  • set_status_pagepost_status_noteTurn the status page on, and post an incident and its fix.
  • run_pipelinereview_pipeline_runsave_pipelineStart, approve, retry, build, and schedule pipelines.
  • find_discussionsadd_pain_point_topicSearch for new threads and turn a pain point into a topic.
  • write_articleapprove_articlesave_content_profileWrite, edit, approve, and schedule articles, and set the writing schedule.
  • track_keywordsrun_seo_auditfind_link_ideasTrack searches and competitors, and run SEO audits and reviews.
  • set_ai_questionsrun_ai_visibilityChoose the buyer questions and run an AI visibility check.
  • update_propertychoose_connection_sitedisconnect_sourceRename or remove a property, pick the site after a sign-in, and remove a connection.
  • save_dashboardset_alert_settingsset_digest_settingsDashboards, alerts, and the weekly summary.

Data

Work on the records in your own product, through its adapter.

  • list_resourcesquery_resourceWhat your adapter exposes, like users or orders, and the records in it.
  • update_resource_rowChange a field your adapter lets people edit.
  • run_resource_actionRun one of your adapter's actions, like a refund. Ones that delete or move money need a confirm.

Every tool, with its inputs and the scope it needs, is listed in the public server card. Older names like list_products and add_product still work, so saved prompts keep running.

Everyday prompts, once it is set up

Your agent reads the same numbers the console shows, while it works in your code. Ask in plain words.

agent · all properties

Which of my products needs me today?

get_healthall properties

PropertyNeeds attention
Product twoSearch Console needs reconnecting
Product one1 step waiting for you
Product threeDown 4 min overnight, back up

Two things need you. Product two's Search Console connection needs a fresh sign-in, and Product one has a step waiting for your approval. Product three dropped for a few minutes overnight and is back.

Start the week

How did this week compare with last, across every product?

get_digest

Find quick search wins

Which searches are we close to page one on for Product one?

read_source

Write the next post

Write next Tuesday's post for Product one in our voice.

next_article_topicsubmit_draft

Check AI answers

When people ask AI assistants for a tool like ours, who gets named?

get_ai_visibility

Count pageviews

Add LopeBase pageview counting to this site and check it works.

get_pageview_snippetget_pageviews

Tell users about an outage

We were down this morning. Post an incident on the status page, then mark it fixed.

post_status_note

One prompt. A product set up while you code.

Free for one property. Agent access works on every plan.

Start free

Safe to hand over

An agent works inside limits you set, asks you for what only you can do, and never sees a stored key.

  • It asks, you approveSign-ins, payments, and pasting a key come back to you as a step, tracked until done. Billing, your team, and agent access stay with you.
  • Keys go in, never outNo tool returns a stored secret. Saved keys are encrypted and the console shows only their last four characters.
  • Scopes you chooseSetup, Read, Manage, or Data. Grant Manage and Data only when you want those changes.
  • Tokens that expire1, 7, 30, or 90 days. Revoke one in Settings and it stops on the next request.
  • Checked on every callA token acts as the admin who made it. If they lose admin rights, it stops working.
  • Names, not secretsThe scan reads dependency and file names. It never opens a real .env file or your source code.
  • Totals, not peopleRead tools return search and analytics totals, never visitor or customer records. Only the Data scope reaches your product's own records.
  • On the recordEach change is in the audit log, under the person who created the token.
  • Never posts for youNo tool lets an agent post on Reddit or a forum.

Guides for every source and assistant

Once a source is connected, your agent can read it for every product. Each guide shows what it reads and prompts to try.

An MCP server for each source

Working across all of them? See one MCP server for all your SEO data. New to MCP? Start with asking AI about your website traffic, or see how Google's own Analytics MCP server compares.

Google data in the AI app you use

Questions

How agent setup works, in plain terms.

Something else? Contact us.

Do I need an account first?

No. When you run npx lopebase, the browser step lets you sign in or create a free account, then you confirm the code from your terminal.

Which agents work with it?

npx lopebase finds Claude Code, Codex, Gemini CLI, opencode, and Cursor Agent. By hand, any agent or editor that can use a remote MCP server over HTTP works, including Claude.ai and ChatGPT on plans that allow custom apps.

What if no agent is found on my computer?

Run npx lopebase prompt, or answer "None, print the prompt" when it asks, and paste the setup prompt into whichever agent you use.

What does the scan read?

Dependency names from your manifests, which config files exist, your public site address, and the variable names in example env files. It asks first, and it never opens a real .env file or your source code.

Can an agent see my secrets?

No. An agent can save a key you give it, but it can never read one back. Saved keys are encrypted, and the console only ever shows their last four characters.

Can I limit what an agent can do?

Yes. Each token carries the scopes you pick: Setup, Read, Manage, or Data. A Read token can look but never change anything. Tokens, scopes, and tools lists what each one allows.

How do I stop an agent?

Revoke its token in Settings. It stops working on the next request. Managing agent tokens has the details.

Is agent access included in every plan?

Yes. Agent tokens work on every plan, including Free. A few manage tools, like a pipeline step that uses AI or a discussions search, spend AI credits as they do in the app.

Hand the setup to your agent.

Free for one property. No card. One command in your product's folder.

  • Agent access on every plan
  • Keys go in, never out
  • Revoke a token any time