> ## Documentation Index
> Fetch the complete documentation index at: https://areyoufoundbyai.com/docs/llms.txt
> Use this file to discover all available pages before exploring further.

# Are you found by AI? MCP tools: full reference

> Every Are you found by AI? MCP tool with its parameters, on the demo, a customer token and an agency connection: get_visibility, get_answers, get_context, get_crawler_access, get_schema_evidence, get_fix_plan, request_rescan and more.

The Are you found by AI? MCP server gives your AI assistant tools it can call mid-conversation. The public demo lists 20 tools: 19 read tools and one action, `request_rescan`. A customer's own token lists more: 30 tools, or 39 when the account owner has turned on editing for the site's connection. Those two counts come from the service source with the current release settings, not from a live customer token, so ask your assistant to list its tools to see your own. The extra tools are listed [below](#more-tools-on-a-customer-token). Every tool is scoped to the token you connect with, so no tool takes a `url` or domain parameter: the token decides which business the answers are about.

<Note>
  **No account yet?** Point your assistant at the demo endpoint (`https://areyoufoundbyai.com/mcp/demo`) to try all 20 tools against our own live monitoring of areyoufoundbyai.com. No key and no signup. On the demo, `request_rescan` returns a worked example of its response and queues nothing.

  In Claude Code: `claude mcp add --transport http found-by-ai https://areyoufoundbyai.com/mcp/demo`
</Note>

Read every result with its measurement date, method and coverage. Missing or unavailable evidence is reported as unknown, never as a negative result.

The evidence tools (`get_crawler_access`, `get_ai_traffic`, `get_agent_view`, `get_benchmark`, `get_share_of_voice`, `get_cited_queries`, `get_source_profile`, `get_citation_sources`, `get_schema_evidence` and `get_regional_visibility`) read saved measurements or existing caches only. They make no provider or browser call and consume no scan allowance. Missing or failed evidence stays unavailable, not zero. Quoted source text in their results is untrusted data.

## The 20 tools on every monitor token and the demo

### 1. `get_visibility`

The latest AI Visibility and AI Readiness scores (each out of 100), with the previous week's scores, the separate off-site Footprint score, and the subscores. Citability, E-E-A-T, technical, schema and platform compose Readiness; Footprint sits beside it, not inside it.

Being named in the answer prose counts in full toward your visibility score. A positive result recorded only as a list entry, citation or source title counts half. Older results without this distinction keep their recorded scoring basis. The AI Visibility score is an older weighted diagnostic score, not a percentage of AI recommendations.

**Parameters:** none.

***

### 2. `get_question_trajectories`

Every tracked buyer question with its measured history: how many of the 7 engines named the business on each measured day. Use it to see, question by question, where visibility is won or lost.

**Parameters:** none.

***

### 3. `get_personas`

The 2 to 4 buyer personas inferred from the business's measured questions, described the way AI models would describe each buyer type, with the exact tracked questions each persona asks.

**Parameters:** none.

***

### 4. `get_mentions`

New pages on the web that mention the business, when a mention check has been run.

| Parameter | Type | Description |
| - | - | - |
| `days` | number *(optional)* | Look-back window in days, 1 to 90 (default 30) |

***

### 5. `get_rivals`

The competitor names the answer engines actually gave in the latest scan when they did not name this business.

**Parameters:** none.

***

### 6. `get_cited_queries`

Questions naming or citing a competitor: a reverse lookup over saved measurements. Give a competitor's brand name or domain and get the buyer questions where the engines named them, which engines, how often, when they were last seen, and who else was named on those questions. It also reports where that domain was cited as a source. Coverage is the questions we have measured, so a thin result means we have not asked those questions yet, never that the competitor is absent from AI.

| Parameter | Type | Description |
| - | - | - |
| `competitor` | string *(required)* | A brand name or domain, for example `Rocketlane` or `rocketlane.com` |
| `engine` | string *(optional)* | Only questions where this engine named them |
| `since` | string *(optional)* | ISO date; only measurements on or after it |
| `limit` | integer *(optional)* | Maximum questions, default 25, maximum 100 |

***

### 7. `get_schema_evidence`

The saved structured-data comparison: initial HTML against rendered JSON-LD, with read dates, incomplete states, excerpts and limitations. It reads stored evidence only and makes no provider request. Page through the findings with `offset` and `limit`.

| Parameter | Type | Description |
| - | - | - |
| `offset` | integer *(optional)* | Where to start in the findings; default 0 |
| `limit` | integer *(optional)* | How many findings to return, default 20, maximum 40 |
| `chars` | integer *(optional)* | Maximum characters per initial or rendered excerpt, default 3000, 120 to 12000 |

***

### 8. `get_regional_visibility`

Saved regional naming observations: areas, buyer questions and engine naming observations, with retained samples and read coverage. Unknown and failed reads are kept distinct from "not named". It makes no fresh provider requests.

**Parameters:** none.

***

### 9. `get_answers`

The exact answer each engine gave to each tracked buyer question in the latest deep measurement: engine, model, date, whether the business was named and on how many samples, live web search or model memory, the competitors named in that answer, the web searches the engine ran before answering (fan-out), and the question's Google demand. This is the receipt behind every score.

| Parameter | Type | Description |
| - | - | - |
| `offset` | integer *(optional)* | Question offset from `nextOffset`; default 0 |
| `question` | string *(optional)* | Substring of a tracked question, case-insensitive |
| `engine` | string *(optional)* | One of ChatGPT, Perplexity, Gemini, Claude, Grok, DeepSeek, Google AI Overviews |
| `limit` | integer *(optional)* | Maximum question rows, default 12, maximum 25 |
| `chars` | integer *(optional)* | Maximum characters per answer, default 4000, maximum 12000 |

<Note>
  **How results are labelled.** Named: your business is named in the answer. Cited: your website is linked in the answer's sources. Listed: your name appears in a list, heading or source title rather than in the prose. Found via search: a page appears in recorded search results, which on its own is not a citation or recommendation. These can overlap. Unavailable is not a negative result.
</Note>

***

### 10. `get_share_of_voice`

Saved share of voice, from saved measurements only. Two measurements, kept apart. First, your share of mentions in your own tracked answers against the rivals the engines named there. Second, web mention share: how often a web-scale index of ChatGPT answers mentions you against the brands we compare you with, which is not a rival list. Use `get_rivals` for who the engines name instead of you.

**Parameters:** none.

***

### 11. `get_context`

The complete weekly context document: scores, every tracked question with its verbatim answer status, mentions, and what to do next. It is the same content as the downloadable `context.md`, so an assistant can load the whole picture in one call.

**Parameters:** none.

***

### 12. `get_fix_plan`

The prioritised fix list from the latest deep scan: the specific changes that move the business up in AI answers, highest impact first.

**Parameters:** none.

***

### 13. `get_citation_sources`

Saved citation sources: the domains AI engines actually cite in answers matching this category, most-cited first, with whether this business appears on each.

**Parameters:** none.

***

### 14. `get_crawler_access`

Saved AI crawler access: whether the business's robots.txt allows or blocks each documented AI crawler from OpenAI, Anthropic, Google, Perplexity, Apple, Microsoft, Meta and Common Crawl, split by what each one is for: search, model training, or fetching a page because a user asked. It reads stored evidence only, never a fresh fetch of robots.txt, and quotes the matching robots.txt rule for each.

It also names the vendors this checker cannot assess, so an agent does not read silence as permission.

**Parameters:** none.

***

### 15. `get_source_profile`

A saved profile of any domain the answer engines cite, from our measurement corpus: how often our measurements saw engines read it and across how many businesses and categories, what the site says it is (title and description from our crawl of its homepage), the co-read pack it travels in (sources the engines read together), and, where our crawler found one, the page where a business gets listed on it. A thin result means the engines rarely cite it in what we have measured so far.

| Parameter | Type | Description |
| - | - | - |
| `source` | string *(required)* | The source domain, for example `canstar.com.au` or `yelp.com` |

***

### 16. `get_post_brief`

Everything an AI needs to draft social posts that move AI visibility, assembled from this week's measured data: fresh third-party mentions to anchor on, the exact buyer questions the engines answer without naming the business (and who they name instead), the domains the engines actually read, and the entity rules that make a post retrievable. It returns a drafting brief, never generated copy: the drafting happens in your AI, in the business's own voice.

**Parameters:** none.

***

### 17. `get_ai_traffic`

Saved AI-referred traffic: the visitors AI engines sent to the business's site, as recorded by the site's own beacon and already saved. It does not start a refresh.

| Parameter | Type | Description |
| - | - | - |
| `refresh` | boolean *(optional)* | This saved-read surface does not start refreshes. Use the separately capped refresh action |
| `confirmed` | boolean *(optional)* | Listed in the tool schema without a description |

***

### 18. `get_agent_view`

The saved agent-browser observation of the site: how a headless agent browser read it (full, partial or blank). A site that renders blank to agents is invisible to agentic AI regardless of content quality. It returns the saved observation and does not start a new read.

| Parameter | Type | Description |
| - | - | - |
| `refresh` | boolean *(optional)* | This saved-read surface does not start refreshes. Use the separately capped refresh action |
| `confirmed` | boolean *(optional)* | Listed in the tool schema without a description |

***

### 19. `get_benchmark`

The measured category benchmark, from saved measurements: where the business sits against every other measured business in its category: rank, percentile, and the distribution above and below.

**Parameters:** none.

***

### 20. `request_rescan`

The demo's one action. It queues a fresh deep measurement now instead of waiting for the next scheduled scan, using one of the plan's capped on-demand re-measures. It runs the measurement and changes nothing else. Results land in a few minutes; then read `get_visibility` again.

**Parameters:** none.

<Warning>
  On a token account, `request_rescan` queues nothing and spends nothing: it replies that the token price needs to be reviewed and confirmed in your workspace. Use `quote_token_run` and `confirm_token_run`, or the workspace itself, to run a paid check. On a subscription that keeps its earlier terms, on-demand rescans are capped at 5 per rolling 7 days. A rescan is refused when the freshest scan is under 20 minutes old, and only one can be queued at a time. If the server refuses, wait for the cooldown or the next scheduled check rather than retrying, and never call `request_rescan` again while polling for a rescan already in progress. On the demo endpoint it returns a worked example and queues nothing.
</Warning>

## More tools on a customer token

The demo does not list any of these. The counts and names below come from the service source with the current release settings (token billing and the client workspace on). We have not listed them against a live customer token, so treat your own assistant's tool list as the final word.

**Three saved-only reads**, no parameters (23 with the 20 above):

| Tool | What it returns |
| - | - |
| `get_balance` | The site's existing rolling on-demand scan allowance, queued status and cooldown. It does not show the token balance; use `get_token_balance` for that |
| `get_weekly_brief` | The saved weekly brief |
| `get_grants` | The saved grants source check |

**Seven workspace reads** (30 in total):

| Tool | Parameters | What it returns |
| - | - | - |
| `get_token_balance` | none | The account's shared token balance and expiry dates. Needs account billing authority |
| `get_token_run` | `id` | The saved state and result of one token run for this site |
| `list_reports` | `offset` | Private author reports for this client, plus the available templates |
| `get_report` | `id`, `revision`, `includeDeleted` | One saved report, or one of its saved versions |
| `list_deleted_reports` | `offset` | Deleted reports that can still be restored during their 30-day recovery period |
| `list_report_versions` | `id`, `offset` | The preserved versions of one report |
| `get_brief` | none | The saved client brief: goals, positioning, audience, key pages and writing preferences |

**Nine workspace actions**, listed only when the account owner has turned on editing for the site's connection (39 in total). Each needs `confirmed: true`, which records that the person agreed:

| Tool | What it does |
| - | - |
| `quote_token_run` | Returns a ten-minute price quote for a run without spending. Show the price to the person before confirming |
| `confirm_token_run` | After the person agrees to the quoted price, reserves the tokens and queues that exact run. A retried request key is not charged twice |
| `set_brief` | Replaces the saved client brief |
| `save_report` | Saves a new client report |
| `update_report` | Replaces a report's content at its current revision |
| `delete_report` | Removes a report; it can be restored for 30 days |
| `restore_report` | Restores a deleted report during its recovery period |
| `restore_report_version` | Makes an earlier version current again, keeping history |
| `regenerate_report` | Attaches the client's latest saved evidence to a report as a new version, without a new scan |

There is no schedule tool on the connection. You choose how often checks run (weekly, monthly or whenever you like) in your signed-in workspace.

## Agency network catalogue

An agency connection (a network link or a per-staff scoped key) reaches a separate catalogue, served as `found-by-ai-fleet`: the network tools below, plus saved-evidence reads for each permitted site. Each per-site tool takes a `domain` that must be one of the sites the connection can reach, as returned by `list_sites`.

| Tool | Parameters | What it returns |
| - | - | - |
| `get_network_summary` | none | The whole network in one read: average AI visibility and readiness, the week's biggest movers, and how many sites are monitored |
| `list_sites` | none | Each site with its latest visibility and readiness, week movement, verdict, tracking wiring and monitoring status, worst performers first |
| `get_site_answers` | `domain`, `offset`, `question`, `engine`, `limit`, `chars` | The verbatim answers behind one site's score: which engines named the business, the sentence each engine gave, who was named instead, and the searches the engines ran first |
| `get_site_rivals` | `domain` | The businesses the engines put in front of buyers instead of this site, plus names that recur across recent measurements |
| `get_site_citations` | `domain`, `limit` | For each tracked question, whether the engines named the business or only used one of its pages as a source. Named and cited are counted separately |
| `get_site_scores` | `domain` | Latest and previous visibility, readiness, verdict and subscores for one site |
| `get_site_mentions` | `domain`, `days` | New pages on the web that mention one site, when a mention check has been run |
| `get_site_fix_plan` | `domain` | The prioritised fix list from one site's latest deep scan |
| `get_site_personas` | `domain` | The buyer personas measured and edited for one site; cards a human pinned or wrote are marked |
| `get_site_post_brief` | `domain` | A drafting brief for social posts from one site's measured data |
| `run_check` | `domain` | Queues a capped on-demand run for one site and returns a run reference, not a result |
| `get_run` | `run` | Given the run reference from `run_check`, says whether that run has produced a measurement yet, and reports its scores once it has |

Alongside those network tools, the catalogue carries saved-evidence reads for each permitted site, each taking `domain`: `get_site_crawler_access`, `get_site_ai_traffic`, `get_site_agent_view`, `get_site_benchmark`, `get_site_share_of_voice`, `get_site_cited_queries`, `get_site_source_profile`, `get_site_citation_sources`, `get_site_schema_evidence`, `get_site_regional_visibility`, `get_site_weekly_brief` and `get_site_grants`. It also has `get_balance` (taking `domain`) for a site's existing scan allowance, and the setup reads `get_site_setup`, `get_edit_history` and `get_question_history`. In the service source, the agency catalogue also carries the workspace tools above, each taking `domain`; the actions appear only for a connection allowed to make changes.

Per-staff keys are scoped and revocable. See [Scoped keys](/docs/guides/scoped-keys).

## Example prompts once connected

After adding the MCP server to your AI assistant, ask questions like these in plain language:

```text theme={null}
@found-by-ai What is my current AI visibility score, and what moved since last week?
@found-by-ai Show me the verbatim answers ChatGPT gave when buyers asked about my category.
@found-by-ai What should I fix first to improve my AI readiness score?
@found-by-ai Which competitors is AI naming instead of me, and on which questions?
@found-by-ai Which third-party sites do the engines cite in my market, and am I on them?
```

Your assistant turns each question into the right tool call and answers from your live measurement data.

## Tool permissions at a glance

* **Read tools** (19 on the demo; on a customer token, also `get_balance`, `get_weekly_brief`, `get_grants` and the seven workspace reads): `get_visibility`, `get_question_trajectories`, `get_personas`, `get_mentions`, `get_rivals`, `get_cited_queries`, `get_schema_evidence`, `get_regional_visibility`, `get_answers`, `get_share_of_voice`, `get_context`, `get_fix_plan`, `get_citation_sources`, `get_crawler_access`, `get_source_profile`, `get_post_brief`, `get_ai_traffic`, `get_agent_view`, `get_benchmark`.
* **Action** (1 on the demo): `request_rescan`, which re-runs your own measurement inside your plan's cap and touches nothing else. On a token account it queues nothing and points you to the workspace.
* **Workspace actions** (9, only when the account owner has turned on editing): `quote_token_run`, `confirm_token_run`, `set_brief` and the six report actions.

These tools answer for the business behind a monitor's token (or the demo). An agency connection uses the network catalogue above instead, where `run_check` is the action.

The MCP connection cannot change your plan or your billing settings. Without editing turned on, its only side effect is `request_rescan`. With editing on, `confirm_token_run` spends tokens from your shared balance, but only after the person has seen and agreed to the quoted price, and the brief and report actions save content in your workspace.


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.