Common questions

Straight answers, in the order people actually ask them. If you don't see yours, ask in the Discord or in Discussions.

Does this work with WordPress?

Yes, if you host WordPress yourself. Connect it once from Settings with an application password - the sort WordPress generates for you under Users, then Profile, then Application Passwords - and finished articles are posted straight to the site: cover image uploaded, internal links to your existing pages inserted, meta description and FAQ markup in place. Nothing to install, no plugin, no repo.

Two things to know. The user you connect needs the Editor role, or it can write drafts but not publish them. And if a security plugin has switched off the WordPress API entirely, the connect screen will say so by name rather than failing later.

Sites on wordpress.com rather than your own hosting aren't supported yet - you'll find out at the connect screen, before anything is charged.

The other way in is a GitHub repo, where articles arrive as pull requests you review: Next.js, Astro, Hugo, Jekyll, anything whose source lives in git. Wix, Squarespace, Shopify, Webflow, Framer and Ghost have neither door. See what you need before installing.

Do I need a coding agent?

No. The ordinary Claude app at claude.ai drives this on any paid Claude plan. You add DispatchSEO as a custom connector under Settings, then Connectors, and your Claude does the research and the writing on your own subscription, then hands the finished article to us to check, finish and publish. Nothing to install, no terminal, and no key of yours ends up with us.

A coding agent still does more: it can build interactive tools, open its own pull requests, and run overnight with nobody watching. The dashboard's Connect your AI screen has both sets of instructions.

ChatGPT can't connect yet, and the Gemini app has no way to connect an outside tool at all. See Do I need a Claude subscription? for which plans cover which path.

Why does Claude keep asking me to allow DispatchSEO?

That is claude.ai's own safety prompt for tools that change something (propose an idea, hand in an article), not an error on our side. Choose Always allow when it asks and it stops asking for that tool. Reads (checking the site, search results, your queue) are marked read-only and normally don't prompt at all.

What do I actually do after connecting the Claude app?

Home shows a "Your next step" list for Claude-app sites. In order: let Claude ask its three questions (paste the setup sentence), ask it for ideas, approve the ones you like on the Queue screen (nothing is written until you do), ask it to write, then watch the Drafts screen go Checking, Ready to publish, Posted, Live. After the first article you only repeat the last two.

Can the Claude app publish to a GitHub site?

Yes. When your site's source lives in a GitHub repo, the article your Claude hands in is committed and opened as a pull request by DispatchSEO itself rather than by an agent on your machine - merged on its own in automatic mode, or waiting for your click in semi-automatic. So a code-built site doesn't need a coding agent either, and the Drafts screen links each pull request and shows when it is merged and live.

Do I need to pay for anything?

Not to run DispatchSEO itself. The free tier (Search Console only) costs $0 forever and self-hosting has no feature limits - see Is the self-hosted version limited? below. You bring your own AI - a paid Claude plan for the Claude app, a Claude subscription for Claude Code, an OpenAI key for Codex, or a Cursor plan's API key for Cursor (next question); beyond that, DispatchSEO never bills you for AI usage - it isn't a per-token service on top of your plan.

What does it cost to run?

The keyword-data tiers stack, and you decide how far up to go:

Free mode finds keywords you can win; paid mode also knows which ones are worth winning. Full breakdown on Google Search Console and Day to day.

There's one more cost outside DispatchSEO entirely: the pipeline runs as GitHub Actions in your own repo, on your own GitHub account, which is free for your first two or so sites and roughly $5/site/month to GitHub beyond that. See GitHub Actions costs for the real numbers and how to avoid a surprise pause.

Do I need a Claude subscription? Which plan?

Some paid Claude plan, if Claude is your AI at all - Codex works too, on an OpenAI API key instead of a subscription (see Use Codex instead), and so does Cursor, on its own plan's API key (see Use Cursor).

For the Claude app path: any paid plan can add a custom connector, and letting Claude repeat the run on a schedule of its own needs Pro or Max.

For the Claude Code path: Claude Code isn't part of the free Claude plan. Pro, Max, Team, and current Enterprise plans all include it; Pro is the cheapest way in if you're starting from nothing. See Install Claude Code for the full plan table and the five-minute setup. For the Codex path there's no subscription at all - OpenAI meters each run against your key, so the account needs credit on it. For the Cursor path, any plan can mint the key; builds draw on the plan's included usage, and the free pool is small enough that a nightly schedule realistically wants a paid plan.

Can I run it on my laptop?

Yes, for a test drive. DispatchSEO works on schedules, and schedules only run while the machine is on - a laptop that sleeps overnight is only running part time. Content builds catch up on their own once it wakes, and Search Console stats mostly catch up too, but daily rank checks are sampled live once a day: a day the machine slept through is a gap in that chart forever, with no way to backfill it. None of this shows up as a false alarm, though - missed schedules are delays, not failures, and the health checks know the difference between "this broke" and "the machine was off." For daily use we recommend something that stays on - a ~$5 VPS, a Raspberry Pi, or a desktop that never sleeps. See Install on your computer and Install on a VPS.

How is this different from SEObot / Outrank / Ahrefs?

Most SEO content tools learn about your product by crawling your homepage - a shallow proxy for something your coding agent already knows in depth, from working in your actual codebase. DispatchSEO doesn't try to replace that knowledge with its own crawler or its own writer model. It's a backend for the agent you already have: MCP tools holding the queue, the keywords, the rankings, and the Search Console data, plus crons that keep that state fresh. The research and the writing happen in your own agent - Claude Code on your subscription, Codex on your OpenAI key, or Cursor on its plan's key - DispatchSEO is state, scheduling, and an approval gate around it, not a competing content engine.

Will Google penalize AI-written content?

Google's own guidance is that velocity and authorship aren't the signal - low-value, mass-produced pages are, whoever wrote them. DispatchSEO's guardrails target exactly that risk: a flat cap of one guide a day per site (no burst publishing), a deterministic sameness check that compares each new guide against your site's own back catalogue rather than just competitors' (which is as far as most commercial SEO tools check), a live SERP check before writing so the angle is actually winnable, and - unless you turn on full auto mode - a human approval step before anything ships. None of that is a guarantee. It's the same discipline a careful human writer would apply, applied consistently every time.

How long until I see results?

SEO is slow, and DispatchSEO's dashboard says so in plain language instead of pretending otherwise. The honest shape of it:

A real site can run ahead of or behind this - it's the typical schedule, not a promise - and the dashboard's Home page tells you which stage you're actually in, computed from your own data instead of a countdown timer.

Can I manage more than one site?

Yes. One deployment manages any number of sites - each project gets its own MCP token, its own queue, its own data, and its own settings. Add the next one from the dashboard the same way you added the first. See Connect your site.

What happens if I stop using it?

Nothing is locked in. Your published content already lives in your own repo as ordinary commits - it stays exactly as published whether DispatchSEO keeps running or not. Uninstalling removes the dashboard, the schedules, and DispatchSEO's own database; it never touches your site's repo. See Uninstalling cleanly.

Is the self-hosted version limited vs the cloud one?

No feature limits either way. Everything the paid cloud does, the self-hosted repo does today, in your own accounts, at $0 beyond whichever paid data tier you choose. The cloud version sells convenience, not capability: bundled keyword data on one bill, a one-click Google connection instead of the service-account ritual, and no machine of your own to keep running.

Can I edit or reject what the agent writes?

Yes. Every guide or tool arrives as a normal pull request against your repo - review it, request changes, edit it directly, or close it, same as any other PR. In semi-automatic mode (the default) nothing merges without you pressing Merge; in automatic mode a page merges itself once its checks pass, but your repo's normal git history still lets you revert it. See Day to day.

What model does it use? Can I change it?

On Claude Code: whatever your own Claude Code is already configured to use. DispatchSEO doesn't pin a model - it runs your agent, on your plan, with whatever default or manually chosen model your session already has. Change it the same way you'd change it for any other Claude Code work.

On Codex: scheduled builds default to gpt-5. Model availability is per-account (some keys 404 on models they can list), so there's an override: set the SEO_CODEX_MODEL variable in your repo's Actions variables, or CODEX_MODEL in your .env on a Docker install.

On Cursor: scheduled builds run Cursor's auto model, and on a free plan that's the only option - pinning a named model via the SEO_CURSOR_MODEL repo variable (or CURSOR_MODEL in .env) needs a paid plan, and fails the run with "Named models unavailable" without one.

Where does my data live?

Self-hosted: entirely in the Postgres container the install creates on your own machine, reachable only over Docker's internal network - never Supabase, never our servers, aside from a once-a-day anonymous install ping carrying a random id and your version number and nothing else (see Security and your data, which also covers turning it off). Hosted (dispatchseo.com): Supabase, operated by the maintainer. Either way, your site's actual content lives in your own GitHub repo, not in DispatchSEO's database. Full detail on Security and your data.

Is it really open source?

Yes - AGPL-3.0. In one sentence: use it, self-host it, fork it, but if you run a modified version as a service for other people, you have to share that source too.

What if my site doesn't have a blog yet?

Not a blocker. The setup wizard asks, and "not sure" is a valid answer - your coding agent checks your repo during setup and decides what to add. See The setup wizard, step by step.

What happens when my GitHub token expires?

Nothing breaks loudly - one-tap merging (and, on a Docker install using the bundled builder, all content building) quietly stops working on the date you picked when creating it. Come back to the GitHub connection step and paste a fresh token; nothing else needs redoing. See Troubleshooting.

What happens if a scheduled job breaks?

You get a red banner on the dashboard's Home page, and an email if you set one up - never silence. A job that's merely skipped (setup incomplete, or a sleeping laptop) is treated differently from a job that actually failed, and DispatchSEO is careful to only alert on the second kind. See Troubleshooting.