# Scope for AI Agents

> Connect Scope to Claude, ChatGPT, or Cursor. Ask how visible your business is in AI search — then let your agent run checks, draft fixes, and track progress. Included with Growth.

Canonical page: https://scope.online/agents

## How it works

01. **Connect** — One click from your assistant's connector directory — or paste a single link. No code, nothing to install.
02. **Ask** — Plain questions, no dashboard required. Your agent pulls the real numbers from Scope.
03. **Act** — Your agent runs checks, drafts schema and FAQ fixes, and tracks what got done — with your approval.

## Connect your stack

### Connect Claude to Scope
URL: https://scope.online/agents/claude
Add Scope from the connector directory, or paste the endpoint manually.

1. Open Claude → Settings → Connectors → browse the directory and add Scope.
2. Or add it manually with the endpoint below and a restricted Scope API key. Reads are free — actions tell you the cost first.
3. Try: “How visible is my business in AI search?” · “Run a readiness check and draft the top fix.” · “Which competitors beat me this week?”

```
https://scope.online/mcp
```

Example prompts:
- "How visible is my business in AI search?"
- "Run a readiness check and draft the top fix."
- "Which competitors beat me this week?"

### Connect ChatGPT to Scope
URL: https://scope.online/agents/chatgpt
Add Scope from the connector directory, then sign in with your Scope account.

1. Open ChatGPT → Settings → Connectors → add Scope from the directory.
2. Enable Scope in the composer’s tools menu, then sign in with your Scope account when prompted.
3. “Pull my Scope visibility report and summarize what changed this week.”

```
https://scope.online/mcp
```

Example prompts:
- "Pull my Scope visibility report and summarize what changed this week."
- "How many checks do I have left this month?"
- "Draft a fix for my top recommendation."

### Connect Cursor to Scope
URL: https://scope.online/agents/cursor
Add Scope to .cursor/mcp.json, or use the one-click deeplink from your Scope dashboard.

1. Add the JSON below to .cursor/mcp.json (or use the one-click deeplink from your Scope dashboard).
2. Restart Cursor, or reload the MCP servers list, so it picks up the new server.
3. Then in chat: “Run a Scope readiness check on this site and fix the top schema issue.”

```json
{
  "mcpServers": {
    "scope": {
      "url": "https://scope.online/mcp",
      "headers": {
        "Authorization": "Bearer YOUR_SCOPE_API_KEY"
      }
    }
  }
}
```

Example prompts:
- "Run a Scope readiness check on this site and fix the top schema issue."
- "What is my current Scope visibility score?"
- "Test the prompt "best plumber near me" before we track it."

### Connect Claude Code or the CLI
URL: https://scope.online/agents/claude-code
One command in Claude Code, or run the npm package anywhere with an API key.

1. Run the Claude Code command below with your restricted Scope API key.
2. Or run the Scope CLI directly in any shell or CI job — it emits JSON only.
3. Skill-based ecosystems (ClawHub, OpenClaw): install the Scope agent skill (SKILL.md).

```bash
# Claude Code
claude mcp add scope --transport http \
  https://scope.online/mcp \
  --header "Authorization: Bearer YOUR_SCOPE_API_KEY"

# Scope CLI, any environment
SCOPE_API_KEY=YOUR_SCOPE_API_KEY npx --yes \
  --package=https://scope.online/scope-mcp-0.1.0.tgz scope whoami
```

Example prompts:
- "scope whoami"
- "scope get-visibility-report --business-id BUSINESS_ID"
- "scope check-prompt --text "best family dentist in Austin""

### Connect any MCP client
URL: https://scope.online/agents#stacks
Any client that speaks Model Context Protocol works with Scope. The raw contract:

1. Point your client at the Streamable HTTP endpoint below with a Bearer-token Authorization header.
2. A restricted key defaults to read access plus permission to run checks — write access is opt-in.
3. 14 tools, reads are free, actions echo their check cost before spending.

```
Endpoint   https://scope.online/mcp   (Streamable HTTP)
Auth       Authorization: Bearer YOUR_SCOPE_API_KEY
           (restricted key — read + run-checks by default)
Tools      14 · reads free · actions echo checks cost first
```

Example prompts:
- "whoami"
- "get_visibility_report"
- "check_prompt"

## What your agent can do (15 tools)

### See

- `whoami` — Your plan & remaining checks. Returns plan, remaining checks, committed burn, auto-upgrade state, and rate-limit status. (Free, requires read)
- `list_businesses` — Your businesses at a glance. Lists businesses on the account with their latest snapshot and scheduled burn. (Free, requires read)
- `get_visibility_report` — Score, per-engine results, and trend. Returns the overall visibility score, per-engine breakdown, and trend over time. (Free, requires read)
- `get_prompt_results` — Results for every tracked prompt. Returns paginated, prompt-level measurements and cadence. (Free, requires read)
- `get_citations` — The sources AI cites — and your gaps. Returns cited sources across engines and competitor citation gaps. (Free, requires read)
- `get_competitors` — Competitors & share of voice. Returns the measured competitor set and share-of-voice signals. (Free, requires read)
- `get_recommendations` — Prioritized fixes & expected impact. Returns prioritized fixes with their expected impact. (Free, requires read)
- `get_readiness_score` — How AI-ready any site is. Scores a website’s readiness for AI visibility; a limited anonymous check is available with no login. (Free, requires read)

### Do

- `add_business` — Add a business — free. Idempotently adds a business by normalized domain. (Free, requires business:write)
- `run_scan` — Run tracked prompts across engines. Runs every tracked prompt across all supported engines. (1 check per tracked prompt, requires scan:run)
- `check_prompt` — Test one prompt — exactly one check. Tests a single prompt across engines without tracking it. (Exactly 1 check, requires scan:run)
- `track_prompt` — Start monitoring a prompt. Commits a prompt to monthly or weekly monitoring for a business. (Free now — adds committed burn, requires scan:run)
- `untrack_prompt` — Stop monitoring a prompt. Removes a tracked prompt from a business, reducing committed burn. (Free, requires scan:run)

### Fix

- `generate_fix` — Draft fixes & content — never touches your site. Drafts schema, FAQ, or content from a recommendation; never writes to the site. (Free, requires read)
- `complete_recommendation` — Mark a fix done — see the before & after. Marks a recommendation complete with an optional note, for before/after attribution. (Free, requires business:write)

## Example prompts

- **Weekly review:** "Every Monday, pull my Scope visibility report, compare it to last week, and summarize what changed and why."
- **Fix my readiness:** "Run a readiness check on my site, pick the highest-impact fix, and draft it for me."
- **Competitor gap:** "Who does ChatGPT recommend instead of me, and which sources give them the edge?"
- **Test before tracking:** "Test the prompt 'best family dentist in Austin' across engines before we track it."
- **Citation audit:** "List the citations AI uses when it mentions my business — and the ones I'm missing."
- **Draft my FAQ:** "Draft FAQ answers for the questions AI gets asked about businesses like mine."

## Trust & control

- **Restricted keys** — Your agent starts with permission to look and run checks — nothing more unless you explicitly allow it.
- **Revocable access** — Disconnect an agent at any time from your Scope settings — access ends immediately.
- **Private by design** — Agents only see what your account allows — and nothing is ever published to your site without you.

## Pricing

Agent access is included with Growth (from $89/month). Starter does not include agent access. See https://scope.online/pricing.

## FAQ

**What is MCP?**
MCP (Model Context Protocol) is the open standard that lets AI assistants like Claude and ChatGPT securely connect to tools. Scope runs an MCP server — connect it once and your assistant can read your visibility data and act on your plan.

**Which agents work with Scope?**
Claude, ChatGPT, Cursor, Claude Code, OpenClaw, and any client that speaks MCP. If your assistant supports connectors or MCP servers, it works with Scope.

**What does agent access cost?**
Nothing extra. Agent access is included with Growth (from $89/mo). Your agent's actions draw from the same monthly pool of checks as the app — one plan, one source of truth.

**What are checks?**
Checks are the unit Scope uses for actions that do work — running a readiness check or testing a prompt. Reports and reads don't spend checks, and your agent always tells you the cost before spending any.

**Can my agent break something?**
No. Today agents draft fixes — schema, FAQ content — and you apply them. Nothing touches your site without your approval, and you can revoke access anytime from your Scope settings.

**Can my agency use one agent for many clients?**
Yes. Connect once with an agency account and your agent can switch between client businesses, run weekly reviews, and draft fixes per client.

**Do I need to be technical?**
No. Connecting is one click from your assistant's connector directory. If you can ask a question in chat, you can operate Scope.

## Endpoints

- MCP endpoint: https://scope.online/mcp
- OAuth issuer: https://scope.online/oauth
- MCP/CLI package: https://scope.online/scope-mcp-0.1.0.tgz
- Tool reference: https://scope.online/agents/tools (markdown: https://scope.online/agents/tools.md)
- Agent skill: https://scope.online/agents/skill (markdown: https://scope.online/agents/skill.md)
- Security & permissions: https://scope.online/agents/security
- llms.txt: https://scope.online/llms.txt