What can Claude Code do for marketing?
Claude Code can do the hands-on half of growth work: fix SEO and AI-readiness issues in your codebase, research and draft articles, plan topic clusters and run reviews of what is working. What it lacks on its own is the data and the priorities, which is the part LogNorm supplies.
Without a source of truth, Claude Code guesses which page to fix and which keyword matters. Connected to LogNorm, it reads your site audit, Search Console data, keyword research, competitors and AI answers, and works the moves your team already ranked. Our guide to Claude Code for marketing covers the workflows; this page covers the setup.
How do you connect Claude Code to LogNorm?
Paste one sentence into Claude Code and click Allow in your browser. Claude Code reads LogNorm's instructions and runs the commands itself.
- Ask Claude Code to connect
Paste
Connect to LogNorm: follow https://lognorm.com/connect.md. Claude Code adds the remote MCP server and installs the LogNorm skill. - Authenticate
Type
/mcp, choose lognorm, then Authenticate. LogNorm opens a consent screen in your browser with your workspace and website filled in. - Allow and join
Press Allow and join team. Claude Code joins with a name of its own and the editor or contributor role, never above yours.
- Restart once
Restart Claude Code so the tools and the skill load. With
--scope user, LogNorm is available in every project.
Setting it up by hand takes two lines: claude mcp add --transport http --scope user lognorm https://lognorm.com/api/mcp, then the skill download shown in Agents and MCP. Codex, Cursor and Claude on desktop or the web connect the same way; see the connect page.
What should you run first?
Start with this week's plan. LogNorm ships eight workflow prompts that appear in Claude Code as slash commands, plus starter prompts on the Agents page with your website and week filled in.
| Command | What Claude Code does |
|---|---|
/mcp__lognorm__weekly_growth | Works this week's plan end to end: team requests first, then fixes, content and AI visibility, with sub-agents |
/mcp__lognorm__fix_audit | Fixes the top audit moves (three by default) in the repo and verifies them on the live site |
/mcp__lognorm__write_post | Takes the top unclaimed content move to a review-ready draft with brief, images and cover |
/mcp__lognorm__plan_topic | Builds a hub-and-spoke plan for a topic and turns the articles into moves |
/mcp__lognorm__ai_visibility | Reviews AI answers, tunes tracked prompts and turns learnings into moves |
/mcp__lognorm__growth_review | Runs a growth review and posts findings as a strategist session |
/mcp__lognorm__requests | Picks up Ask your agent requests your team left on the dashboard |
/mcp__lognorm__setup | Fills in positioning, competitors, topics, AI prompts and key product facts on a new workspace |
Which permissions should Claude Code get?
Give it the permissions for the jobs you want it to do and switch off the rest. Reading is always on; everything else is on by default and can be changed at connect time or later on the Agents page.
| Permission | What it allows |
|---|---|
| Read this website's LogNorm data | Moves, the weekly plan, audits, keywords, AI visibility, competitors, content and the Company Brain. Always on. |
| Work on moves | Claim moves, move them through their stages, comment and start sub-agents |
| Write drafts | Save articles as drafts and send them to review. Publishing stays with your team. |
| Re-check fixes | Re-run the audit check for an issue once a fix is live, for a few credits |
| Add to the Company Brain | Save documents, pages and memories the team and LogNorm's writer use |
| Run research | Re-crawl the site, run audits, check AI visibility or find keywords, within your spend cap |
Customize on the consent screen also sets the website, the role and when Claude Code leaves the team if idle: after 1, 7 or 30 days. When the same app reconnects an agent you already allowed, it reconnects without showing the screen again.
Can Claude Code run marketing work in parallel?
Yes. Claude Code can start sub-agents and register each one with LogNorm, so a fixer, a writer and a researcher work the same week side by side. Each sub-agent appears under its parent on the dashboard and claims its own moves.
Claims are atomic, so two sub-agents never pick the same move, and a claim lapses two hours after the last action. When a sub-agent finishes, any move it still holds goes back to the parent. Requests and comments are rate limited per agent, so a runaway loop can't flood your team's notifications.
How does Claude Code read your marketing data?
Instead of dozens of read tools, Claude Code writes a little JavaScript against a typed, read-only lognorm client, and LogNorm runs it in a sandbox. It filters and joins inside the code and brings back only what it needs, which keeps its context small.
Every method is limited to the agent's website. The sandbox has no network, no files and no timers, and allows up to 60 LogNorm calls and 40 seconds per run. Text from third parties, such as competitor pages or AI answers, is screened for prompt injection before Claude Code sees it.
What does the trail look like?
Everything Claude Code does shows up live on your dashboard, credited to it. Your team reads its work where the work lives, not in your terminal history.
- Claims and a live status line while it works.
- Comments on each move with its plan, decisions, changed files and pull request links.
- Drafts in review, with notes beside the text on what it changed and which sources it chose.
- Sub-agents shown under their parent, such as Panda · Pip.
- Every action in the activity log under its name.
What does it cost to run Claude Code with LogNorm?
Claude Code's own reasoning and writing run on your Claude plan and cost no LogNorm credits. You don't need an Anthropic API key in LogNorm either.
Work LogNorm runs for it, such as a keyword lookup, a research run or re-checking a fix, costs what it costs in the app, within your spend cap. You can switch Run research off at connect time if you want Claude Code to work only from scheduled data. People still approve and publish every draft. See pricing.
Frequently asked questions
Can Claude Code be used for marketing?
Yes. It can fix SEO issues in your codebase, research and write articles, and analyse data. Connected to a source of truth like LogNorm, it works on ranked priorities instead of guesses.
Do I need to be a developer to use Claude Code for marketing?
You need to be comfortable in a terminal. If you're not, connect Claude on desktop or the web instead; it does the same LogNorm work, but can't change your site's code.
Is Claude Code better than Codex for marketing work?
Both connect to LogNorm the same way and use the same tools. Our Claude Code vs Codex comparison covers where each one is stronger.
Can Claude Code publish to my CMS?
No. It saves drafts and sends them for review. A person approves and publishes to your CMS or merges the pull request.
How do I disconnect Claude Code?
Remove it from the Agents page in LogNorm. Its access ends at once and its claimed moves go back to the team.
Keep reading
- All solutionsSee how different teams use LogNorm to find what holds growth back, rank the work and ship it with their AI agents, from solo founders to growth teams.
- Claude Code for marketingRun SEO and growth from your terminal.
- Agents and MCPCommands, permissions and every tool.
- Agents (MCP)Your coding agent as a teammate.
- AI marketing agentThe agents-on-the-team model.
- What is MCP?The Model Context Protocol for marketers.
- Claude Code vs CursorFor marketing, SEO and growth work.