Self-hosting DispatchSEO

The full guide lives on the docs site: dispatchseo.com/docs. It's the same step-by-step content, one page per task, and it's also served by your own instance at /docs once you're running - so it works offline too. This file is just the map.

The short version - run from any plain folder (not inside your website's repo; it creates its own dispatchseo folder). On a VPS, ssh in first and run this on the server, not on your own computer:

git clone https://github.com/NeoZi12/dispatchseo &&
  cd dispatchseo &&
  sh start.sh

On Windows, paste this version instead - it works in plain PowerShell (Git must be installed) and installs into your user folder no matter where the window opened:

cd ~; git clone https://github.com/NeoZi12/dispatchseo; cd dispatchseo; .\start.cmd

Open the URL it prints (usually http://localhost:4005), choose a dashboard password, and the setup wizard takes it from there. (The VPS guide below has a shorter one-line installer plus your own domain with automatic HTTPS - this block is the plain version that works anywhere.)

The guide, page by page

• Install on your own computer - Docker Desktop, the one command, what's running, the builder, upgrading. Includes the honest laptop-vs-always-on trade-off: builds catch up when a laptop wakes and GSC stats mostly catch up too, but daily rank checks leave permanent gaps, so we highly recommend a machine that stays on (a ~$5 VPS, a Raspberry Pi, or a desktop that never sleeps) for real use - a laptop is fine for a first look, not for day-to-day.
• Install on a VPS - one line installs everything (Docker included), then one DNS record + one .env line puts the dashboard on your subdomain with automatic HTTPS. This is the path we recommend once you're past trying it out.
• The setup wizard, step by step - what each screen asks for and why: your site, the Search Console service account, keyword data (free mode vs DataForSEO), publish mode, one-tap merge, and the two pastes that connect your coding agent.
• Day to day - what using it actually looks like after setup, plus the data tiers.
• The dashboard, page by page - every screen, what it shows, and when to open it.
• Automations and modes - the three modes, all twelve automations, publishing pace, and the quality gates.
• Troubleshooting - what each failure means and how to fix it, symptom by symptom.
• Upgrading and backups - pulling a new image, backing up Postgres, and moving to another machine.

Reference, once you want the details:

• Environment variables - every variable, whether you need it, and where to get the value.
• Schedules and jobs - what runs when, and what you see when something fails.
• MCP tools - every tool your agent can call.
• Security and your data - the auth model, what is stored, and what leaves your machine.
• Architecture - how the pieces fit together.

What you need

• A computer that can run Docker - about 1 GB of RAM. A laptop works for a test drive; something always-on for running it seriously.
• Your website's code in a GitHub repo. Content ships as pull requests, so git-based sites only; WordPress and other database-backed CMSes won't work.
• A coding agent - Claude Code or Codex. Your own agent does the research and the writing, and it also runs the in-stack builder that works while you're away. Claude Code runs on the Claude subscription you already pay for; Codex runs on your own OpenAI API key, which OpenAI meters per run. Pick one on the dashboard's Settings → Coding agent - the builder reads that, so switching needs no reinstall. Its credential goes in one of two places: pasted on Home's "Turn on automatic builds" card, or set as CLAUDE_CODE_OAUTH_TOKEN / OPENAI_API_KEY in your .env.
• Google Search Console access to your site - free, at search.google.com/search-console.

Advanced: environment variables

The wizard stores everything it collects (encrypted) in your own database, so a normal install needs no hand-set variables beyond what start.sh creates. An environment variable always wins over the wizard's stored value.

The full annotated list is on the docs site - Environment variables - and in .env.docker.example as raw comments.

The daily install ping

Once a day your install tells dispatchseo.com two things: a random id generated on your machine at first boot, and the version you're running. That is the entire payload - no domain, no email, no keywords, no site or Search Console data, no tokens, and the id isn't derived from any of them.

It exists because nothing else can answer "how many people actually run this". Clone counts and image pulls measure download attempts, and one person re-running start.sh looks identical to ten separate installs.

To turn it off, put this in your .env and re-run sh start.sh:

DISPATCHSEO_TELEMETRY=off

Nothing else changes - no nag, no degraded features, no reduced support. The code is src/lib/heartbeat.ts if you'd rather read it than take our word for it.