ContextBolt SEO in Muse Code (2026 Setup)

Muse Code is Meta’s terminal coding agent. It runs on Muse Spark 1.3 with a 1M-token context window, edits across large repos, and reads remote MCP servers over streamable HTTP from a settings file. That puts live keyword, SERP, competitor, backlink, and Search Console data beside the pages it is already changing.

ContextBolt SEO is the hosted data layer. You add one block to the settings file, sign in once, and ask Muse Code for the SEO job rather than opening a separate dashboard.

Add the server from settings.json

Muse Code has no add command for MCP servers. You edit ~/.config/muse/settings.json, then sign in. The field names below come from Meta’s extending and automating documentation.

{
  "schema_version": 1,
  "mcp_servers": {
    "contextbolt-seo": {
      "transport": "streamable_http",
      "url": "https://seo.contextbolt.app/mcp",
      "enabled": true,
      "mode": "optional"
    }
  }
}

schema_version is not optional. A settings file without it fails every muse command at startup. streamable_http is the transport for a remote server reached by URL, and it takes no framing field. mode defaults to required, which aborts the run if the server fails to start, so set it to optional for a research tool.

Then connect your account with OAuth.

muse mcp login contextbolt-seo

The browser opens the ContextBolt sign-in page. Check the address is on seo.contextbolt.app, enter your SEO account email and the email code, and choose Allow connection. The shared URL has no secret in it; your credentials are stored by Muse Code, not in the file.

Start muse in a project and type /mcp. It shows the connected servers and the tools each one gives the session. Ask, What SEO tools do you have? and you get live keyword research, difficulty, SERPs, domain and competitor analysis, backlinks, site audits, AI visibility, and Search Console tools.

If Muse Code is not installed yet, it is one command on macOS and Linux, curl -fsSL https://dev.meta.ai/install.sh | bash, and a PowerShell one-liner from install.ps1 on Windows, then muse login. Meta’s Muse Code page lists the plans that include it.

Muse Code and Muse are different surfaces

The names are close, but the setup is not.

• Muse Code runs in your terminal. It reads ~/.config/muse/settings.json, signs in with muse mcp login, and works beside project files. A local server works here.
• Muse is Meta’s personal agent on the web, iOS, Android, and WhatsApp. It runs on a cloud computer of its own, so a server on your machine is invisible to it. It connects to a hosted server through a Custom Connector you ask it to build. The Muse setup is that path, and SEO with Muse is the longer walkthrough.

If the job is recurring research while your laptop is closed, Muse is the shape. If the job sits beside a site or codebase you are actively changing, Muse Code keeps the research and the fix in one terminal session.

What Muse Code can do with live SEO data

The useful shift is from asking for ideas to asking for evidence and an action in the same task.

For an existing site, ask Muse Code to audit it. The audit picks the most important pages itself, 25 by default and 50 at most, one credit per page. Muse Code groups the findings, fixes the title and meta issues in the working tree, runs the project’s checks, and leaves the diff for review. Ask it to re-audit after the deploy and it re-checks the same pages and leads with what changed. muse exec runs that re-check as a one-liner from the shell without opening a session.

For a new page, ask it to expand the topic, pull volume and difficulty, inspect the live top 10, and choose the target that fits the site’s authority. It can then use those results while it writes the page.

For your own numbers, connect Search Console and ask which pages lost clicks against the previous period. Muse Code can open the relevant files, compare each page with the current SERP, and propose the change.

For authority work, ask for the domains that link to competitors but not to you. It filters the result, groups prospects by angle, and saves a shortlist beside the project. Backlink calls cost three credits rather than one.

These are all jobs an SEO MCP server enables. The agent decides what to do, and the server supplies the numbers and current search results it needs to decide from. The longer walkthrough, with the prompts, is in SEO inside Muse Code.

Keep the connection trusted

Meta’s docs say MCP tools run outside Muse Code’s sandbox, as network connections rather than sandboxed shell commands, so connect servers you trust. ContextBolt SEO’s data tools are read-only lookups against your own credit balance. Nothing in it writes to your website or your accounts.

The shared OAuth URL is safe to keep in a settings file. The personal URL from your welcome email is not, because its token sits in the path, so treat that whole URL like a password if you use the compatibility route.

One more label worth reading. Meta’s model page marks the cheaper contributor model as “Used to improve our products” and the standard model as “Not used to improve our products.” If you pay per token on the contributor price, do not route a client’s Search Console data through it.

The same URL also connects to Claude Code, Codex, Cursor, Grok Build, Hermes, and OpenClaw. The SEO for agents hub maps the always-on options, and the connection guide covers reauthentication.