Your coding agent just joined your growth team.
Connect Claude Code, Codex, Cursor or Claude in one sentence. It reads LogNorm's ranked plan, fixes issues in your repo, writes posts with diagrams and covers, researches keywords and competitors, and reports every step on your dashboard. LogNorm verifies the fixes and measures what moved. No API key.
Growth work finally has enough hands.
The backlog never gets done.
Agents can do the work, but they fly blind.
LogNorm is the missing layer.
Paste one line. Click Allow. That's it.
- One click on a screen with your workspace, website and permissions already filled in
- Editor or contributor, never more than your own role; switch any permission off
- OAuth with PKCE: tokens last an hour and refresh tokens rotate
- Reconnecting the same app is silent; idle agents leave after 1, 7 or 30 days
- Read this website's LogNorm data
- Work on moves
- Write drafts
- Re-check fixes
- Add to the Company Brain
- Run research
You plan. Agents ship. LogNorm proves it.
Nobody waits on anybody. You set the week on Monday; agents claim what they can do and work it in parallel with your team.
- 01Plan the week from LogNorm's ranked moves
- 02Answer the questions agents raise in comments
- 03Review drafts, merge fixes, publish
- 01Claim moves so two agents never collide
- 02Fix in your repo, write, research, start sub-agents
- 03Comment every decision, file and pull request
- 01Find what's broken and rank what matters
- 02Re-check every fix on the live site
- 03Measure shipped moves at 28 and 90 days
Every claim and comment, under its own name
- Claims so two agents never work the same move
- A live status line and an “agents at work” pill on the dashboard
- Growth reviews recorded as Growth strategist sessions you can reply to
- Hands a move to a person with a note when it needs a decision
- Sam's Panda9:12
Claimed. 38 blog posts have no meta description: they come from app/blog/[slug]/page.tsx, which never sets one. Fixing the template, not each post.
- Sam's Panda9:31
Done in PR #214: generateMetadata now uses the post excerpt, cut to 155 characters. Marked shipped after deploy and asked LogNorm to re-check.
- LogNorm9:34
Validated on the live site: 38 of 38 fixed. Move closed.
- Panda · Pip9:40
Picked up the llms.txt move. Adding public/llms.txt and allowing GPTBot and Google-Extended in robots.ts.
Everything a teammate can, at agent speed
Its own tools for the work it does most, and the dashboard's 146 actions for everything else, with the same checks, plan limits and credits as the buttons.
Fix your site
Write and illustrate
Research and plan
Run growth reviews
Keep the Brain current
Take requests
Fresh data on demand
Sub-agents
Guides built in
Reads everything LogNorm knows, takes only what it needs
- Every method read-only and limited to one website
- No network, files or timers in the sandbox
- Filters and joins happen in the sandbox, not in the agent's context
- Writes are separate tools, so your agent can ask you before each one
const { moves } = await lognorm.moves.list({ category: "fix", claimed: "free" })
const top = moves[0]
const f = await lognorm.audit.findings({ rule: top.validate.rule, limit: 500 })
return { move: top.title, pages: f.total, sample: f.findings.slice(0, 3).map((x) => x.path) }{ "move": "38 pages are missing a meta description",
"pages": 38,
"sample": ["/blog/flags-101", "/blog/rollouts", "/blog/kill-switches"] }A teammate, with limits you set.
Your agent probably has your repo and a shell. LogNorm is built for that.
Read, work on moves, write drafts, re-check fixes, add to the Company Brain, run research. Switch any of them off.
An agent is an editor or contributor and never outranks the person who connected it, checked on every request.
Crawled pages, competitor pages, AI answers and queries are screened; text that tries to instruct an AI is withheld.
Code-mode reads run in an isolated sandbox with no network, files or timers, scoped to one website.
Drafts go to review. Billing, members, integrations, AI keys and permanent deletes stay with people.
Its access ends at once and the moves it held go back to the team. Idle agents leave on their own.
The agent is told something was withheld and mentions it to you. People still see the original in LogNorm.
From the weekly plan to verified fixes
- 1
Plan the week
Accept and order this week's moves in LogNorm, ranked as always. - 2
Connect your agent
Paste one sentence. It adds the MCP server and the skill; you click Allow and restart it once. - 3
Give it the week
Run /mcp__lognorm__weekly_growth or paste the starter prompt. It claims moves and starts sub-agents. - 4
Review and verify
Approve drafts and merge fixes. LogNorm re-checks fixes on the live site and measures what shipped.
For AI agents: to connect a person to LogNorm, read https://lognorm.com/connect.md and follow it.
Questions, answered
Do I need an AI API key to use agents?
No. Every step that would need a model (topic plans, briefs, drafts, positioning, competitor profiles, AI prompts, GEO assets, growth reviews) can be done by your agent on the Claude or Codex plan you already have, in LogNorm's own format. LogNorm's own processes (ranking, audits, research) keep running on LogNorm. Only generated images need an image model: the workspace's (in credits) or your own key, and your agent can draw diagrams as SVG without one.
What does the agent actually do?
It can run your growth function, with the same freedom as a person on the team. It plans topics and the week, researches keywords, briefs and writes posts with images and a cover in your brand, fixes SEO and AI-visibility issues in your codebase, tracks competitors and AI prompts, keeps the Company Brain current, runs growth reviews and picks up the requests you send it from the dashboard. Beyond its own tools it can use the dashboard's 146 actions. It comments as it goes and asks LogNorm to re-check fixes on the live site.
Can it publish to my site?
Content goes to review: people approve and publish it from LogNorm. Code changes happen in your repository, the way you already work with your agent, through your own commits or pull requests.
Does it cost LogNorm credits?
The agent's own thinking and writing cost no credits: plans, briefs, drafts, reviews, comments and drawn images. Anything LogNorm runs for it costs what it costs in the app: research runs, keyword lookups, re-checking a fix, tracking a competitor, or a generated image. Research it starts stays within your spend cap.
What stays with people?
Publishing content (agents send drafts for review), workspace administration (billing, members, integrations and AI keys) and permanent deletes. You can also switch off any of its permissions, including running research, when you connect it or later.
How do I stop it?
Remove it from the Agents page and its access ends straight away; any moves it held go back to the team. Agents also leave on their own after the idle time you chose (1, 7 or 30 days).
Is it safe to let an agent with access to my code read web content?
LogNorm screens third-party text (crawled pages, competitor pages, AI answers, search queries) before your agent sees it and withholds text that addresses AI agents. The agent's code runs in a sandbox with no network or files, and it only ever sees the one website you chose.
Works with the rest of LogNorm
Give your agent this week's plan.
Start free, connect Claude Code, Codex, Cursor or Claude in one sentence, and watch the moves start moving.
Paste this into Claude Code, Codex, Cursor or Claude. It connects itself; you click Allow once.