Plugin for IDEs and AI agents
Your SEO agent lives in your editor
The agent reads your audit's advice, edits your site's code, marks it done and sees how many clicks you gained. It works on your site and your Search Console, not on generic market data.
/plugin marketplace add swellweb/seoautohub-plugin /plugin install seoautohub@seoautohub
One command in Claude Code. For Cursor, Codex and OpenCode it takes about four lines of config, below.
How it works
From advice to a measured result, without leaving the editor
The agent reads what to change
Titles, H1s and descriptions to improve, with the chosen search and the Search Console numbers (impressions, clicks, position).
It applies the change in your code
It finds the page in your project, shows before and after, and applies it when you confirm. It does not invent text beyond the advice.
It verifies and measures
It marks the change done: the hub re-reads the live page and, after 28 days, compares clicks before and after with a verdict.
"Done" is never taken on trust: the hub checks the published page. If the change is not live, it tells you.
Installation
Connect it in three steps
- Create a token in SEO Automation Hub, Settings, IDE and AI agent access. Pick "Read only" or "Read and actions".
- Store it as an environment variable:
export SEOAUTOHUB_TOKEN="…" - Add the server to your client by copying the block below.
Claude Code
Full plugin: skills and server together
/plugin marketplace add swellweb/seoautohub-plugin /plugin install seoautohub@seoautohub
MCP server only, without skills:
claude mcp add --transport http seoautohub https://seoautohub.com/api/mcp \ --header "Authorization: Bearer $SEOAUTOHUB_TOKEN"
Cursor
~/.cursor/mcp.json
{
"mcpServers": {
"seoautohub": {
"url": "https://seoautohub.com/api/mcp",
"headers": {
"Authorization": "Bearer ${env:SEOAUTOHUB_TOKEN}"
}
}
}
}
Codex
~/.codex/config.toml
[mcp_servers.seoautohub] url = "https://seoautohub.com/api/mcp" bearer_token_env_var = "SEOAUTOHUB_TOKEN"
OpenCode
opencode.json
{
"mcp": {
"seoautohub": {
"type": "remote",
"url": "https://seoautohub.com/api/mcp",
"headers": {
"Authorization": "Bearer {env:SEOAUTOHUB_TOKEN}"
}
}
}
}
Tools
What the agent can do
| Tool | Permission | What it does |
|---|---|---|
list_sites |
read | Lists your sites with audit status and progress. |
list_suggestions |
read | What to change on each page, with the reasoning and Search Console numbers. |
get_results |
read | Clicks before and after, verdict and the effect of each change made. |
run_audit |
write | Starts a new audit of the site. |
mark_suggestion_done |
write | Declares a change applied: the hub verifies it on the live page. |
Limits: one audit per hour per site and one at a time per account. The in-depth AI analysis covers the pages most viewed on Google; the number depends on your plan.
What you need
Requirements, no surprises
Professional plan or higher
The plugin is free and open source. Letting the agent read your data is included from the Professional plan, and during the free trial.
A personal token
You create it and revoke it whenever you like. The value is shown once, and a read-only token cannot change anything.
Search Console connected
Click and impression numbers come from your Google account. Without it the audit works but measurements stay empty.
Questions
Frequently asked questions
Is the plugin free?
Yes, the plugin is open source under the MIT licence. Reading your account's data needs the Professional plan or higher, free trial included.
Can the agent change things without asking me?
Not on your site: code changes are made by your client under its own confirmation rules. A read-only token cannot start audits or mark changes as done.
What data does the agent see?
Only your account's: sites, advice, results. Never other users'. You can revoke the token at any time.
How is it different from an SEO data connector?
Data connectors give the agent market numbers to explore. This one works on your site: advice, change, verification on the live page and click measurement with your own Search Console.