io.github.tarang8811/getintel

Official25 tools

by tarang8811

Manage GetIntel AI visibility (GEO) from your own AI agent.

io.github.tarang8811/getintel — Model Context Protocol (MCP) Server

The io.github.tarang8811/getintel MCP server provides capabilities to manage GetIntel AI visibility (GEO) from your own AI agent. Its purpose is to integrate GetIntel GEO visibility management into an agent workflow via MCP, using the server as the interface.

🛠️ Key Features

Manage GetIntel AI visibility (GEO) from an AI agent

🚀 Use Cases

Integrate GetIntel GEO visibility management into an AI agent
Centralize control of GetIntel AI visibility through MCP

⚡ Developer Benefits

Lets developers handle GetIntel (GEO) visibility via an MCP server exposed to their agent

⚠️ Limitations

Available details only describe GEO visibility management; no additional capabilities, tools, or constraints are provided

Declared in the repository manifest (tools.json). Not verified against a live endpoint.

get_visibility

Get the AI Visibility score (0-100) — how often AI named the brand across the last 7 days of scans (a stable trailing average, not a single day). score is NULL when nothing was measured; 0 means AI answered and named the brand nowhere. Also returns scans_in_window and prompts_observed (the confidence behind the score), the change vs the previous 7-day window, and up to 12 points of history.

Parameters

No parameters.

Raw schema

{
  "type": "object",
  "properties": {},
  "required": []
}

get_ai_answers

See which buyer questions AI search engines were asked and, for each, the RATE the brand was named across the last 7 days of scans (over runs_counted) plus the per-engine state: 'open' (AI named no product at all — a winnable slot), 'absent' (AI named others, not this brand), 'mentioned', 'recommended', 'no_answer', or 'error' (the run failed).

Parameters

No parameters.

Raw schema

{
  "type": "object",
  "properties": {},
  "required": []
}

get_receipts

Get the exact answer each AI engine gave for one buyer question over the last 7 days, and how many of the (non-errored) runs named the brand.

Parameters1

prompt

string

required

The exact buyer question text, as returned by get_ai_answers.

Raw schema

{
  "type": "object",
  "properties": {
    "prompt": {
      "type": "string",
      "description": "The exact buyer question text, as returned by get_ai_answers."
    }
  },
  "required": [
    "prompt"
  ]
}

list_activity

List recent things the agent found or did — signals, approvals, drafts, outcomes — most recent first.

Parameters2

limit	integer	optional	Max rows to return. Default 20.
category	string	optional	Optional category filter.

Raw schema

{
  "type": "object",
  "properties": {
    "limit": {
      "type": "integer",
      "description": "Max rows to return. Default 20."
    },
    "category": {
      "type": "string",
      "enum": [
        "watch",
        "act",
        "outcomes"
      ],
      "description": "Optional category filter."
    }
  },
  "required": []
}

list_missions

List missions the agent has proposed or is running — batches of related opportunities waiting for approval.

Parameters1

state

string

optional

Raw schema

{
  "type": "object",
  "properties": {
    "state": {
      "type": "string",
      "enum": [
        "proposed",
        "approved",
        "executing",
        "completed",
        "cancelled",
        "failed"
      ]
    }
  },
  "required": []
}

get_mission

Read-only preview of one mission: the buyer opportunities it bundles and any drafts already written — WITHOUT running it. Use this to inspect a mission before the irreversible approve_mission.

Parameters1

mission_id

integer

required

Raw schema

{
  "type": "object",
  "properties": {
    "mission_id": {
      "type": "integer"
    }
  },
  "required": [
    "mission_id"
  ]
}

get_competitors

Share-of-voice board: how often each AI engine names the brand vs each competitor over the last 30 days — who's winning, and on which buyer prompts.

Parameters

No parameters.

Raw schema

{
  "type": "object",
  "properties": {},
  "required": []
}

get_prompt_matrix

The per-prompt × per-engine matrix. For each buyer question, each engine's DOMINANT state over the last 7 days (open/absent/mentioned/recommended/no_answer) + the rate behind it, the brand's position when named, plus a visibility count of how many engines named the brand. no_answer means the engine returned nothing to read (failed call, or Google rendered no AI Overview) — it is NOT an open slot.

Parameters

No parameters.

Raw schema

{
  "type": "object",
  "properties": {},
  "required": []
}

get_sources

Which domains the AI engines cite in the brand's category, how often, and whether each is the brand's own / a competitor's / third-party — where to get published to earn citations.

Parameters1

limit

integer

optional

Max sources to return (1-50, default 15).

Raw schema

{
  "type": "object",
  "properties": {
    "limit": {
      "type": "integer",
      "description": "Max sources to return (1-50, default 15)."
    }
  },
  "required": []
}

get_content_gaps

Topic coverage: for each buyer question, whether the brand has a page, whether AI cites it, the recommended action (write / improve / leave), and (for gaps) a page_type hint — commerce_page vs editorial — for ecommerce/local/hybrid brands.

Parameters

No parameters.

Raw schema

{
  "type": "object",
  "properties": {},
  "required": []
}

get_signals

Recent Reddit + X signals the agent surfaced — buyer threads AI already cites where the brand is absent, competitor complaints, and ICP discovery posts.

Parameters1

limit

integer

optional

Max signals to return (1-50, default 20).

Raw schema

{
  "type": "object",
  "properties": {
    "limit": {
      "type": "integer",
      "description": "Max signals to return (1-50, default 20)."
    }
  },
  "required": []
}

get_search_console

Google Search Console performance joined against the tracked prompt set, so you can find pages and queries that rank on Google but that AI never cites. Returns totals (clicks, impressions, ctr, impression-weighted avg_position), a rows array, and ranks_but_never_cited — rows in the top 10 of Google that AI cited zero times. With dimension="page", each Search Console URL is matched exactly against the URLs AI cited (tracking params and www. normalized away). With dimension="query", the keyword is matched to a tracked prompt by normalized substring containment. Errors clearly if Search Console is not connected or the grant was revoked.

Parameters3

days	integer	optional	Trailing window in days (1-90, default 28).
dimension	string	optional	"query" for keywords, "page" for URLs. Default "query".
limit	integer	optional	Max rows to return, ranked by impressions (1-100, default 25).

Raw schema

{
  "type": "object",
  "properties": {
    "days": {
      "type": "integer",
      "description": "Trailing window in days (1-90, default 28)."
    },
    "dimension": {
      "type": "string",
      "enum": [
        "query",
        "page"
      ],
      "description": "\"query\" for keywords, \"page\" for URLs. Default \"query\"."
    },
    "limit": {
      "type": "integer",
      "description": "Max rows to return, ranked by impressions (1-100, default 25)."
    }
  },
  "required": []
}

get_authority

Link authority for the brand's domain: domain_rating, referring_domains, backlinks, and net new referring domains over 7/30 days. Also returns link_opportunities — high-DR link targets and listing directories, ranked by domain rating, with directories the brand has already submitted to filtered out.

Parameters2

limit	integer	optional	Max link targets and directories to return, each (1-50, default 10).
category	string	optional	Optional category filter applied to both dr_targets and directories.

Raw schema

{
  "type": "object",
  "properties": {
    "limit": {
      "type": "integer",
      "description": "Max link targets and directories to return, each (1-50, default 10)."
    },
    "category": {
      "type": "string",
      "description": "Optional category filter applied to both dr_targets and directories."
    }
  },
  "required": []
}

get_brand

Does AI know this brand, and would it recommend it when asked — the trust/reputation signal (distinct from get_visibility's raw citation rate, which has no sentiment). Reads the last stored weekly run, not a live probe. Deliberately lightweight — verdicts/labels/scores only, not the full AI answer text; call get_brand_transcript for that.

Parameters

No parameters.

Raw schema

{
  "type": "object",
  "properties": {},
  "required": []
}

get_brand_transcript

The full text get_brand leaves out: the complete prompt/answer/citations for every turn actually run (findability, blind ask, named follow-up) — real, untruncated AI answer text, not a summary.

Parameters1

engine

string

optional

Optional — one of openai/perplexity/gemini_direct/google_aio. Omit to get all engines' transcripts.

Raw schema

{
  "type": "object",
  "properties": {
    "engine": {
      "type": "string",
      "description": "Optional — one of openai/perplexity/gemini_direct/google_aio. Omit to get all engines' transcripts."
    }
  },
  "required": []
}

get_crawlability

Can AI reach and read the site at all — the technical prerequisite layer underneath both get_visibility and get_brand. 12 pass/warn/fail checks (SSL, robots.txt, firewall, response time, llms.txt, sitemap.xml, JS-only rendering, noindex, snippet blocking, entity schema, FAQ/HowTo schema, title/meta) run against the homepage plus a second page from the sitemap.

Parameters

No parameters.

Raw schema

{
  "type": "object",
  "properties": {},
  "required": []
}

get_usage

This month's MCP budget: reads and drafts used / limit / remaining, and when it resets.

Parameters

No parameters.

Raw schema

{
  "type": "object",
  "properties": {},
  "required": []
}

get_prompt_history

Per-prompt movement: which buyer questions you WON, LOST, or held between the current 7-day window and the window a full 7 days earlier, with the mention rate in each.

Parameters

No parameters.

Raw schema

{
  "type": "object",
  "properties": {},
  "required": []
}

list_drafts

List drafts the agent has written — outreach emails, blog pitches, llms.txt / schema drafts, Reddit replies. Summary rows; call get_draft for the full text.

Parameters2

state	string	optional
limit	integer	optional	Max drafts to return (1-50, default 20).

Raw schema

{
  "type": "object",
  "properties": {
    "state": {
      "type": "string",
      "enum": [
        "draft",
        "approved",
        "copied_to_clipboard",
        "sent",
        "discarded"
      ]
    },
    "limit": {
      "type": "integer",
      "description": "Max drafts to return (1-50, default 20)."
    }
  },
  "required": []
}

get_draft

Get the FULL content of one draft — the actual article / llms.txt / schema / reply text the agent generated, so you can ship it.

Parameters1

draft_id

integer

required

Raw schema

{
  "type": "object",
  "properties": {
    "draft_id": {
      "type": "integer"
    }
  },
  "required": [
    "draft_id"
  ]
}

approve_mission

Approve one mission. This runs the ENTIRE batch headlessly right now — drafts get written; outbound sends still require the user's send mandate to already be on. Not reversible once it executes.

Parameters1

mission_id

integer

required

Raw schema

{
  "type": "object",
  "properties": {
    "mission_id": {
      "type": "integer"
    }
  },
  "required": [
    "mission_id"
  ]
}

dismiss_mission

Dismiss a proposed mission without running it.

Parameters1

mission_id

integer

required

Raw schema

{
  "type": "object",
  "properties": {
    "mission_id": {
      "type": "integer"
    }
  },
  "required": [
    "mission_id"
  ]
}

replan

Ask the agent to re-scan current signals and propose fresh missions right now, instead of waiting for the weekly scan.

Parameters

No parameters.

Raw schema

{
  "type": "object",
  "properties": {},
  "required": []
}

approve_event

Approve one individual finding (not a whole mission) — writes its draft/article. Use for a single item rather than a batch.

Parameters1

event_id

integer

required

Raw schema

{
  "type": "object",
  "properties": {
    "event_id": {
      "type": "integer"
    }
  },
  "required": [
    "event_id"
  ]
}

dismiss_event

Dismiss one individual finding.

Parameters1

event_id

integer

required

Raw schema

{
  "type": "object",
  "properties": {
    "event_id": {
      "type": "integer"
    }
  },
  "required": [
    "event_id"
  ]
}

README.md6 KBView on GitHub

GetIntel MCP Connector

GetIntel is an AI visibility (GEO) platform — it tracks how often ChatGPT, Perplexity, Gemini, and Google AI Overviews name your brand in buyer conversations, finds the gaps, and drafts the fixes. This connector lets your own AI agent (Claude, or any MCP client) read that data and act on it directly, in conversation.

Server: https://app.getintel.ai/mcp (remote, streamable-http)
Auth: OAuth 2.1 (Dynamic Client Registration + PKCE) — no pasted API keys
Registry: io.github.tarang8811/getintel on the Official MCP Registry
Privacy policy: https://app.getintel.ai/privacy
Requires: an active GetIntel account

This repository holds the connector's public documentation and tool manifest. The server implementation lives in GetIntel's main application repo.

Connecting

In your MCP client, add a remote server with URL https://app.getintel.ai/mcp.
Complete the OAuth flow — you'll be redirected to sign in / authorize with your GetIntel account and choose which scopes to grant (read, act).
The client discovers available tools automatically via tools/list — a read-only token only sees the read tools below.

Each brand in your account gets its own token, scoped so an agent connected for one brand can never see another brand's data — see tools.json for the full schema.

Tools

Every call is scoped to the authenticated user/brand. read tools never mutate anything.

get_visibility - AI Visibility score (0-100) — trailing 7-day citation rate, with trend and history.
get_ai_answers - Buyer questions × engine grid — where the brand is open, absent, mentioned, or recommended.
get_receipts - The exact AI answer for one buyer question, across engines.
list_activity - Recent things the agent found or did, most recent first.
list_missions - Proposed/active batches of related opportunities awaiting approval.
get_mission - Read-only preview of one mission before approving it.
get_competitors - Share-of-voice vs named competitors over 30 days.
get_prompt_matrix - Per-prompt × per-engine state matrix.
get_sources - Domains AI engines cite in your category — where to earn citations.
get_content_gaps - Per-question coverage: do you have a page, does AI cite it, what to do.
get_signals - Reddit/X threads AI cites where you're absent, or competitor complaints.
get_search_console - Google Search Console performance joined against tracked AI prompts.
get_authority - Domain rating, referring domains, and link-building opportunities.
get_brand - Does AI trust/recommend the brand by name — the reputation signal.
get_brand_transcript - Full untruncated prompt/answer/citation text behind get_brand.
get_crawlability - Can AI crawlers actually reach and read the site (12 pass/warn/fail checks).
get_usage - This month's MCP read/draft budget and when it resets.
get_prompt_history - Buyer questions won/lost vs the prior 7-day window.
list_drafts - Drafts the agent has written (outreach, articles, llms.txt, replies).
get_draft - Full content of one draft.
approve_mission - Approve and execute a mission's whole batch. Not reversible once run.
dismiss_mission - Dismiss a proposed mission without running it.
replan - Re-scan current signals and propose fresh missions now.
approve_event - Approve one individual finding, writing its draft.
dismiss_event - Dismiss one individual finding.

Read

Tool	Description
`get_visibility`	AI Visibility score (0-100) — trailing 7-day citation rate, with trend and history.
`get_ai_answers`	Buyer questions × engine grid — where the brand is open, absent, mentioned, or recommended.
`get_receipts`	The exact AI answer for one buyer question, across engines.
`list_activity`	Recent things the agent found or did, most recent first.
`list_missions`	Proposed/active batches of related opportunities awaiting approval.
`get_mission`	Read-only preview of one mission before approving it.
`get_competitors`	Share-of-voice vs named competitors over 30 days.
`get_prompt_matrix`	Per-prompt × per-engine state matrix.
`get_sources`	Domains AI engines cite in your category — where to earn citations.
`get_content_gaps`	Per-question coverage: do you have a page, does AI cite it, what to do.
`get_signals`	Reddit/X threads AI cites where you're absent, or competitor complaints.
`get_search_console`	Google Search Console performance joined against tracked AI prompts.
`get_authority`	Domain rating, referring domains, and link-building opportunities.
`get_brand`	Does AI trust/recommend the brand by name — the reputation signal.
`get_brand_transcript`	Full untruncated prompt/answer/citation text behind `get_brand`.
`get_crawlability`	Can AI crawlers actually reach and read the site (12 pass/warn/fail checks).
`get_usage`	This month's MCP read/draft budget and when it resets.
`get_prompt_history`	Buyer questions won/lost vs the prior 7-day window.
`list_drafts`	Drafts the agent has written (outreach, articles, llms.txt, replies).
`get_draft`	Full content of one draft.

Act

Tool	Description
`approve_mission`	Approve and execute a mission's whole batch. Not reversible once run.
`dismiss_mission`	Dismiss a proposed mission without running it.
`replan`	Re-scan current signals and propose fresh missions now.
`approve_event`	Approve one individual finding, writing its draft.
`dismiss_event`	Dismiss one individual finding.

act tools that send outbound email or publish content still require the account's explicit send/publish mandate — a connected agent can never grant itself that autonomy.

Support

Issues or questions: tarang@sidepanda.com

Install

Tools (25)

get_visibilityget_ai_answersget_receiptslist_activitylist_missionsget_missionget_competitorsget_prompt_matrix

Remote endpoint

Streamable HTTP

Hosted server - connect over the network, no local install.

url

https://app.getintel.ai/mcp

Add to VS Code Add to Cursor

claude_desktop_config.json

json

{
  "mcpServers": {
    "getintel": {
      "command": "npx",
      "args": [
        "-y",
        "mcp-remote",
        "https://app.getintel.ai/mcp"
      ]
    }
  }
}

Desktop config is stdio-only; this bridges via mcp-remote. Native remote: Settings > Connectors.