Your first website in ClicksPal

ClicksPal connects SEO research to the work in your website repository. Start with one website and one page: connect your tools, review the evidence, build a change, then check it before publishing.

This guide follows the desktop app’s setup, website settings and content pipeline. You will need a local Git repository for a website you can edit and publish.

Install the app and connect an AI

Choose your platform on the download page. On Mac, open the DMG and drag ClicksPal into Applications. On Windows, run the installer. The download page identifies the Windows build as an unsigned beta.

Open ClicksPal and activate your license. The setup guide starts with Connect an AI. Choose Claude Code, Codex, OpenCode or Ollama and follow the installation or sign-in controls. The selected provider must pass its connection test before you continue. Your AI account and its usage limits remain separate from ClicksPal.

Install Git, Node.js/npm and your website’s dependencies on the same computer. The app needs the website’s local build tools to check changes; installing ClicksPal does not install your project’s dependencies.

Connect the website and its repository

Choose Add website, enter the public URL and select the root of its local Git repository. Review the product description, audience and verified product facts. These tell build agents what the product actually does. Correct an empty or outdated description before asking for content.

Open Repository commands and paths in the website settings. Check these against your project:

• The article, tool and marketing page directories point to the places your website actually serves.
• The validation command runs your website’s checks. For a project that provides it, this might be npm run build.
• The deploy command matches your hosting setup. Use git push only if a push really deploys this repository.
• Shared code directories and public supporting files contain only the places agents should be allowed to change.

For your first build, uncheck Publish automatically when a build finishes. This lets you inspect the output before it goes live. Keep Autopilot off while you learn the workflow.

ClicksPal inspects the repository and records its reference pages, components and build decisions. If it asks about a missing product fact or a publishing convention, answer the question instead of asking it to guess. You can also select a connected website as a template reference; its layout can be reused without copying that product’s claims or credentials.

Connect research data and run an audit

In Account settings, connect your own DataForSEO account for keyword and search data, then set a research budget. Google Search Console is optional and adds your website’s observed query and page performance. Research provider charges are separate from your ClicksPal license and AI subscription.

Run an SEO audit for the connected website. Read the findings and keyword opportunities alongside their evidence. Missing measurements stay unknown; they are not a reason to claim there is no demand or no traffic.

The audit reads your website and prepares a roadmap. It does not publish a website change. Check proposed topics against your product facts and existing pages before you decide what to build.

Pick one task from the content calendar

Open Content pipeline. Validated opportunities and actionable fixes from the audit can appear on the calendar automatically. Review their briefs, keep useful opportunities, and defer or reject ideas that do not fit. You can move a scheduled task or pause it.

Start with one task whose purpose is easy to verify: an article answering a question your product can address, an improvement to an existing page, or a small tool with clear inputs and outputs. Read the task’s brief before starting it.

The execution model used for builds is separate from the model selected for SEO audits. Choose it for the work you are about to run, without changing the audit’s settings.

Build, answer questions and review

Build the selected task. ClicksPal plans, edits an isolated Git worktree and runs the configured checks. The task records its progress and review results. A built task is not the same as a live page.

Open its preview and check the parts a build command cannot judge: factual claims, links, headings, forms and how the page reads at a narrow width. Confirm that it uses your website’s existing components and patterns.

If a task is blocked or waiting for input, read the actual reason and give it a concrete answer. For example, identify the correct article directory or explain a product limitation that the draft got wrong. Keep credentials in the connection settings or your website’s environment, not in task feedback.

Publish and check the live page

When you are satisfied with the built change, choose Publish. ClicksPal commits the built work, runs the website’s deploy command and checks the resulting page live. Watch the publication result and open the live URL yourself before moving to the next task.

A failed deployment or live check needs attention even when the build passed. Read the failure before retrying. You can request a revision to a published page as another task, so the correction goes through the same build and review flow.

Once this first cycle works, you can choose a publication policy and Autopilot schedule that suit your workflow. Scheduled work requires the app and computer to be running; a sleeping or shut-down computer cannot build pages.

Watch AI usage and connect your agent

Account settings → AI shows ClicksPal calls today and over seven days, measured tokens and provider-reported quota windows. Unknown provider quota is shown as unknown. These numbers help explain a pause; they do not replace your provider’s account limits.

You can explicitly choose a secondary AI for capacity limits. A partially executed writable turn pauses instead of blindly replaying the same actions with a second provider. Read the task’s state before continuing it.

If you prefer working through Claude Code, Codex or another agent, connect it from the app’s MCP page. The agent uses the same websites, tasks and calendar as the window. The MCP documentation includes setup instructions and the actual tool reference.