GitHub Copilot has two surfaces that speak MCP, and they are configured in completely different places. This page covers the cloud coding agent, the one you assign an issue to and that opens a pull request on its own. For SEO tools inside the editor, see the VS Code setup.
The cloud agent is the more interesting of the two for SEO, because it turns research into something that happens automatically as part of your workflow rather than something you remember to do.
What changes when the agent has SEO tools
Assign Copilot an issue that says “add a comparison page for X” and, without SEO tools, it writes whatever seems reasonable. With ContextBolt SEO connected, it can check what the page should target, look at who currently ranks, and write against real data before it opens the PR.
The pattern that pays off fastest is auditing inside pull requests. Point the agent at the pages a PR changed, have it run page_audit on each, and it leaves the findings as a review comment. SEO review stops being a separate task someone forgets.
The repository MCP configuration
Open your repository, go to Settings, then Copilot under Code & automation in the sidebar, then Coding agent. The MCP configuration section takes a JSON block:
{
"mcpServers": {
"contextbolt-seo": {
"type": "http",
"url": "$COPILOT_MCP_CONTEXTBOLT_SEO_URL"
}
}
}Store the full MCP URL as an agent secret rather than pasting it inline. The configuration supports secret and environment variable substitution drawn from the Agents secrets on the repository, using the same syntax style as GitHub Actions. Because your ContextBolt token lives inside the URL path, the whole URL is the thing worth protecting. Check GitHub’s current docs for the exact secret naming convention your organization expects, since the prefix requirements have shifted between releases.
You need repository admin rights to see this panel. On an organization-owned repository, confirm your org’s MCP policy allows third-party servers before adding one.
The OAuth limitation, and why it does not bite here
GitHub’s coding agent supports remote MCP servers but does not support servers that authenticate using OAuth. That rules out a number of hosted MCP services.
ContextBolt SEO carries its token in the URL path instead, so it connects cleanly. It is worth knowing this constraint if you are evaluating several SEO MCP servers at once, because it quietly disqualifies some of them from this surface.
The agent also supports MCP tools only, not resources or prompts. Every ContextBolt SEO capability is a tool, so nothing is missing.
Watch your first few sessions
An agent chaining research across a whole issue can spend more lookups than a single question in an editor. The six research tools cost one lookup each, the three backlink tools cost three. Check the session logs on your first few tasks to learn what a typical issue costs, then size that against your 1,000 monthly lookups.
The full tour of tools and workflows is in the ContextBolt SEO guide, and the background on how an SEO MCP server works sits alongside it.
How to connect ContextBolt SEO to GitHub Copilot
-
Subscribe and get your MCP URL
Subscribe to ContextBolt SEO ($35/month for 1,000 lookups). Your private MCP URL arrives by email and looks like https://seo.contextbolt.app/mcp/YOUR_TOKEN. Keep it private, like a password.
-
Store the URL as an agent secret
In your repository, open Settings, then Environments, and add a secret holding your full MCP URL under the Agents secrets used by the coding agent. Storing the URL rather than pasting it into the config keeps your token out of a settings screen your whole team can read.
-
Open the MCP configuration panel
In the repository, open Settings, then under Code & automation in the sidebar choose Copilot, then Coding agent. Scroll to the MCP configuration section. You need repository admin rights to see this panel.
-
Add the server and save
Paste a JSON block with mcpServers containing contextbolt-seo, type http, and your URL referenced from the secret. Save, then assign Copilot an issue and check the agent's session log to confirm the SEO tools loaded.
Example prompts for ContextBolt SEO + GitHub Copilot
Once connected, try asking GitHub Copilot:
Assign Copilot an issue for a new page and it runs keyword_research and serp_overview first, so the draft is written against a real target.
Calls page_audit on each changed URL and leaves the findings as a PR comment, turning SEO review into part of code review.
Runs competitor_keywords, then files the gaps as issues so the content backlog builds itself.
Uses the free Search Console tools to read your real data and summarize what moved, without spending lookups.
What you can do with ContextBolt SEO in GitHub Copilot
Give the cloud coding agent live keyword and SERP data
Audit changed pages automatically inside a pull request
Score keyword difficulty before a page gets written
Find the keywords a competitor already ranks for
Turn keyword gaps into issues on your backlog
Connect your own Google Search Console, read-only and free
Works alongside any other MCP servers on the repository