Cited Cli MCP Server

Cited CLI

Command-line interface for the Cited GEO (Generative Engine Optimization) platform. See how often ChatGPT, Gemini and Claude cite your business, get recommendations, and generate the fix files, all from the terminal or a script.

For AI agents

• Free audit, no login: call the free_audit MCP tool (or run cited audit free example.com) to see how AI answers 5 buyer questions about a site, and who it cites instead.
• MCP server: https://mcp.youcited.com/mcp (streamable HTTP, OAuth login on first use). Also on PyPI as cited-mcp for local stdio.
• One-command setup: cited agent setup connects Claude Code, Cursor and Codex (and prints the Claude Desktop connector steps). Add --dry-run to preview.
• No account yet: cited agent signup (or the agent_signup tool) creates an account and API key without a login; give your human the claim link and code.
• Docs: the search_docs tool, or cited docs search "<question>", searches the public Cited docs without a login.

Installation

Homebrew (recommended)

brew install youcited/cited/cited

Use the full name: Homebrew 7 refuses formulae from third-party taps it doesn't trust, and installing by full name trusts just this formula. Tapping first and then running brew install cited now fails with "Refusing to load formula ... from untrusted tap".

Upgrading an older Homebrew install

Homebrew 7 only loads formulae from trusted taps. If you installed with brew tap before, trust the formula once (safe to re-run), then upgrade:

brew trust --formula youcited/cited/cited
brew upgrade cited

uv or pipx

With Python 3.11+:

uv tool install cited-cli   # or: pipx install cited-cli

Upgrade later with uv tool upgrade cited-cli (or pipx upgrade cited-cli). To run it once without installing: uvx --from cited-cli cited audit free example.com.

From source (development)

git clone https://github.com/YouCited/cited-cli.git
cd cited-cli
./scripts/dev-install.sh

Quick Start

cited audit free example.com     # free first look, no account needed
cited login                      # opens the browser
cited status                     # who you are, credits, businesses and the next step
cited business setup --website example.com
cited --help

The workflow section below walks through a paid audit, recommendations and fixes.

---

Authentication

# Browser login (default — opens youcited.com login page)
cited login

# Specific OAuth provider
cited login --provider google
cited login --provider microsoft
cited login --provider github

# Password login (non-interactive / CI)
cited login --email you@example.com --password "Secret123!"

# Check status
cited auth status

# Log out
cited logout

Password registration is a two-step email-verification flow:

cited register --email you@example.com --name "Your Name" --password "Secret123!"
# → Check your inbox, paste the verification URL when prompted

---

MCP Server

cited-mcp and the hosted server at mcp.youcited.com give AI assistants Cited's tools: set up a business, draft questions, run audits, get recommendations and generate fixes.

Quick setup

cited agent setup             # configure every AI client found on this machine
cited agent setup --dry-run   # preview the changes first
cited agent setup --client cursor --yes

cited agent setup merges into your existing config (it never replaces an existing cited entry), backs up any file it changes, and supports --client claude-code|claude-desktop|cursor|codex|all, --dry-run, --yes and --json. The manual steps for each client follow.

Claude Code

claude mcp add --transport http --scope user cited https://mcp.youcited.com/mcp

Then run /mcp in Claude Code and choose Authenticate. For Cited slash commands, skills and the geo-analyst agent, also install the plugin:

/plugin marketplace add YouCited/cited-plugins
/plugin install cited@youcited-cited-plugins

Claude Desktop (Custom Connector — recommended)

No local install required. Claude connects to Cited's hosted MCP server directly:

1. Open Claude Desktop → Settings → Customize > Connectors
2. Click "+" → "Add custom connector"
3. Enter URL: https://mcp.youcited.com/mcp
4. Click Add

Your browser will open for authentication on first use. See Anthropic's connector guide for details.

Requires: Claude Pro, Max, Team, or Enterprise plan.

Claude Desktop (Developer Config — alternative)

If Custom Connectors aren't available, use the developer config with mcp-remote:

{
  "mcpServers": {
    "cited": {
      "command": "npx",
      "args": ["mcp-remote", "https://mcp.youcited.com/mcp"]
    }
  }
}

Add to Settings → Developer → Edit Config, then restart Claude Desktop.

Prerequisite: Node.js (LTS) must be installed for npx.

Claude Desktop (local stdio)

{
  "mcpServers": {
    "cited": {
      "command": "uvx",
      "args": ["cited-mcp"],
      "env": {
        "CITED_TOKEN": "your-jwt-token"
      }
    }
  }
}

With no env vars, call the login tool to sign in with a browser, or agent_signup to make an account. To reuse a CLI login, get the token with cited login then cited auth token.

Cursor

Add to ~/.cursor/mcp.json (global) or .cursor/mcp.json in your project root — or run cited agent setup --client cursor:

{
  "mcpServers": {
    "cited": {
      "url": "https://mcp.youcited.com/mcp"
    }
  }
}

Open Settings (Cmd+,) → MCP, click Login next to cited, and verify it shows a green status. Use tools in Agent mode.

Fallback (older Cursor versions without remote-server support) — proxy through mcp-remote, which needs Node.js:

{
  "mcpServers": {
    "cited": {
      "command": "npx",
      "args": ["mcp-remote", "https://mcp.youcited.com/mcp"]
    }
  }
}

Codex

codex mcp add cited --url https://mcp.youcited.com/mcp
codex mcp login cited

Or add [mcp_servers.cited] with url = "https://mcp.youcited.com/mcp" to ~/.codex/config.toml (cited agent setup --client codex does this for you).

One set of tools everywhere

The tools are defined once, in the Cited API. The hosted server at mcp.youcited.com, the local cited-mcp server and this CLI all run the same tools (GET /agent/tools lists them, POST /agent/tools/{name} runs one), so they behave the same. New tools show up in cited-mcp without a release; hosted clients may need to reconnect the Cited connector to refresh the cached list.

cited-mcp adds local login and logout tools. cited tools list prints the live list and cited tools call NAME '{...}' calls any tool with JSON arguments. See packages/mcp/README.md for the MCP server's details.

Example

Once connected, ask your assistant something like:

"Set up Acme Corp from acme.com, draft the buyer questions, and tell me what an audit costs."

It chains the tools: setup_business → draft_questions → quote_audit, then run_audit once you approve the price.

Pricing: pay-as-you-go credits

No plans or tiers: every command and tool works on every account, and work is paid with credits (1 credit = $1). cited audit quote shows what an audit costs, and cited status shows the balance and prices. cited credits buy prints a Stripe checkout link for a person to open; nothing is bought until the balance changes. cited audit run and cited fix generate show the price and ask before spending (--yes skips the question).

---

Global Flags

These flags apply to every command and must be placed immediately after cited:

---

Workflow

Commands that need a business take --business/-b, which is optional when you have one business (or set cited config set default_business_id <id>). Audit commands default to the newest audit.

# 1. A free first look (no account)
cited audit free acme.com

# 2. Sign in (or `cited agent signup` for an agent) and add the business
cited login
cited business setup --website acme.com
cited business setup --id <business_id> --persona "Ops lead: runs a 20-person warehouse"

# 3. Draft the buyer questions the audit asks, then review them
cited audit questions                       # show them, drafting a set if there is none
cited audit questions --add --count 5 --focus "pricing"
cited audit questions -q "Best WMS for small warehouses?" -q "..."   # replace the set

# 4. Price it and run it
cited audit quote
cited audit run                             # shows the price and asks; --yes to skip
cited audit get --wait                      # status, then the report

# 5. What to change
cited recommend list
cited recommend set <item_key> dismissed

# 6. Fix files, then check them live
cited fix generate --all                    # asks before spending
cited fix list
cited fix get <fix_id> --save ./fixes       # write the files to disk
cited fix done <fix_id>                     # after putting it on the site
cited fix verify <fix_id>                   # re-check the live site (free)

# 7. Keep watching
cited audit schedule --every 30
cited audit compare <audit_id>              # against the audit before it
cited business brief

Scripting

Every command takes --json and prints the tool's result unchanged, so the JSON is the same shape an MCP client sees:

BUSINESS=$(cited --json business setup --website acme.com | jq -r '.business.id')
cited --json audit run -b "$BUSINESS" --yes | jq -r '.audit_id'
cited --json recommend list | jq -r '.items[] | "\(.item_key)\t\(.title)"'

cited tools call runs any tool with JSON arguments, for anything the commands don't expose:

cited tools call get_audit '{"limit": 5}'

A command that would spend credits without --yes prints the price instead and exits 0 in --json mode, with "confirmation_required": true in the result.

---

Command Reference

cited login | logout | register | status | version
cited auth login|logout|status|token|register
cited config set|get|show|environments

cited agent setup [--client ...] [--dry-run] [--yes]  ← Connect AI clients to Cited
cited agent signup [--name] [--email] [--free-audit] ← Account + API key for an agent (no login)

cited credits buy [PACK_ID] [--one-audit] [--voucher CODE] [--no-save-card]

cited business setup [--website URL | --id ID] [--name ...] [--persona "Name: desc"] ...
cited business get [BUSINESS_ID]
cited business brief [-b ID]

cited audit free URL [--token TOKEN] [--wait/--no-wait]
cited audit questions [-q ...] [--regenerate] [--add] [--count N] [--focus TEXT] [--job ID]
cited audit quote [-m MODEL ...]
cited audit run [-m MODEL ...] [--engine ...] [--country CC] [--deep] [--yes]
cited audit get [AUDIT_ID] [--answers] [--wait] [--limit N]
cited audit compare AUDIT_ID [--baseline ID]
cited audit schedule [--every DAYS | --off] [--email/--no-email] [--webhook URL]

cited recommend list [--audit ID] [--run ID] [--all]
cited recommend set ITEM_KEY open|done|dismissed

cited fix generate [ITEM_KEY ...] [--all] [--yes]
cited fix list [--state STATE]
cited fix get FIX_ID [--save DIR]
cited fix done|undo|verify FIX_ID

cited docs search QUERY | cited docs get PATH
cited tools list | cited tools call NAME [JSON]
cited mcp serve

---

Configuration

Config is stored in ~/.cited/config.toml:

# Set default environment (saves typing --env dev every time)
cited config set environment dev

# Set default output format (saves typing --json every time)
cited config set output json    # or "text" (default)

# Set a default business (used by commands that accept --business)
cited config set default_business_id <uuid>

# Use an agent API key instead of `cited login` (sent as X-Agent-API-Key on every
# request; `cited agent signup` saves one for you). A login session takes precedence.
cited config set agent_api_key <key>

# Inspect current config
cited config show
cited config environments

Tip: With output set to json, use --text on any command to get human-readable output for that one invocation.

---

JSON Output

All commands write clean JSON to stdout with --json, so they pipe into jq, python3 -c, or any other tool. Errors always go to stderr as {"error": true, "message": "..."} so they never corrupt stdout pipelines.

---

Exit Codes

---

Development

./scripts/dev-install.sh
pytest -v
ruff check src/ packages/
mypy packages/core/src packages/mcp/src src/cited_cli/ --ignore-missing-imports

packages/core/src/cited_core/tools.json is a snapshot of the API's tool list, used when the API can't be reached. Refresh it after the API's tools change:

python scripts/sync_tools_manifest.py            # from api.youcited.com
python scripts/sync_tools_manifest.py --env dev

Releasing

Release and deployment scripts are in a private infrastructure repo. See that repo's README for release and deploy instructions.