# Connect an AI agent to LogNorm

> For AI agents: a person asked you to connect to LogNorm (their growth platform). Follow the steps for the tool you are running in. Add the server and the LogNorm skill (one command each); the person then clicks **Allow** once in their browser and restarts once. No API keys, no packages to install.

LogNorm is a remote MCP server: `https://lognorm.com/api/mcp` (Streamable HTTP, OAuth 2.1 with dynamic client registration and PKCE). Signing in adds you to the person's workspace as a named teammate with access to their website's data, moves and tools.

## 1. Add the LogNorm MCP server and the LogNorm skill

Do both for the tool you are running in. The skill holds LogNorm's playbooks (fix audit issues, write posts, plan topics, run growth reviews) so you use LogNorm well in every session; it contains instructions only, no credentials.

### Claude Code
```bash
claude mcp add --transport http --scope user lognorm https://lognorm.com/api/mcp
curl -fsSL https://lognorm.com/api/agents/skill.zip -o /tmp/lognorm-skill.zip && mkdir -p ~/.claude/skills && unzip -o /tmp/lognorm-skill.zip -d ~/.claude/skills
```
Then tell the person: "Type `/mcp`, choose **lognorm**, then **Authenticate**, and click **Allow** in the browser. Then restart Claude Code once so the LogNorm skill and tools load." (`--scope user` makes LogNorm available in every project.)

### OpenAI Codex (CLI or IDE extension)
```bash
codex mcp add lognorm --url https://lognorm.com/api/mcp
curl -fsSL https://lognorm.com/api/agents/skill.zip -o /tmp/lognorm-skill.zip && mkdir -p ~/.agents/skills && unzip -o /tmp/lognorm-skill.zip -d ~/.agents/skills
codex mcp login lognorm
```
The login command opens the browser; the person clicks **Allow**. Check with `codex mcp list`, then ask them to restart Codex once so the skill and tools load.

### Cursor
Add this to `~/.cursor/mcp.json` (create the file if it doesn't exist; keep other servers in it):
```json
{ "mcpServers": { "lognorm": { "url": "https://lognorm.com/api/mcp" } } }
```
And install the skill where Cursor reads agent skills:
```bash
curl -fsSL https://lognorm.com/api/agents/skill.zip -o /tmp/lognorm-skill.zip && mkdir -p ~/.cursor/skills && unzip -o /tmp/lognorm-skill.zip -d ~/.cursor/skills
```
Then tell the person to open Cursor Settings → MCP and click **Connect** next to lognorm.

### Claude (desktop app or claude.ai)
You can't add connectors or skills yourself here. Tell the person:
1. "Open Settings → Connectors → **Add custom connector**, name it LogNorm, paste `https://lognorm.com/api/mcp`, then click **Connect** and **Allow**."
2. "Download https://lognorm.com/api/agents/skill.zip and upload it in Settings → Capabilities → Skills."

### Any other MCP client
Add a remote (Streamable HTTP) MCP server named `lognorm` with the URL `https://lognorm.com/api/mcp`. It answers 401 with OAuth metadata, so clients that support MCP authorization start the sign-in on their own. If the client supports Agent Skills, install https://lognorm.com/api/agents/skill.zip into its skills folder (the skill index is at https://lognorm.com/.well-known/agent-skills/index.json).

## 2. The person approves (one click)

The browser opens LogNorm. If they aren't signed in, they sign in first. They see the workspace, website and what you'll be allowed to do, and click **Allow and join team**. You join as "<their name>'s <animal>" (for example "Sam's Panda"). If they don't have a LogNorm account yet, they can create one at https://lognorm.com/sign-up and then retry.

## 3. Restart once, then check it worked

After the restart, the `lognorm` skill and tools are available. Call the `search` tool with no query: it returns who you are in the workspace, the website, what data exists and how fresh it is, open requests from the team, and the guides. If you get "Sign in with OAuth to use LogNorm", the approval hasn't finished: ask the person to complete step 2.

## 4. Start working

Tell the person you're connected and offer to start, for example: "Want me to work this week's LogNorm plan?" The server's instructions and `lognorm.guide.get({ id: "start" })` explain everything LogNorm can do. Workflows are also available as MCP prompts (in Claude Code: `/mcp__lognorm__weekly_growth`, `/mcp__lognorm__growth_review`, `/mcp__lognorm__fix_audit`, `/mcp__lognorm__write_post`).

## Notes

- Nothing secret goes in prompts or files: access comes from the OAuth sign-in and can be removed any time in LogNorm → Agents.
- One connection works on one website; connect again to pick another.
- Docs for people: https://lognorm.com/docs/agents