Answer Radar / MCP connector

Answer engine optimization, inside your AI assistant

Measure where ChatGPT, Gemini, Perplexity and Claude recommend a competitor instead of you, and read the sources they cited, without leaving the assistant you already work in.

Answer RadarMCP connector

Run the whole loop from your AI assistant

Answer Radar is not only a dashboard. The same measurement is a connector, so the tool that asks the questions is the one you already work in.

Paste a URL, authorize, done

Add the connector in Claude, Cursor, VS Code or any MCP client, authorize once in the browser, and that is the setup. It is a public OAuth 2.1 client with PKCE, so there is no secret to store and no key to rotate.

  • One URL and a client id, no client secret
  • No API keys, nothing to generate or rotate
  • Works in Claude, Claude Code, Cursor, VS Code, Windsurf and ChatGPT
Claude · Add custom connectorOAuth 2.1

Connector URL

mcp.linkeddit.com/mcp

OAuth Client ID

linkeddit-mcp-claude

Client secret

Not required, public PKCE client

Connect
Nothing to generate or rotateno API keys

The whole loop, from the assistant

65 tools in one server, 37 of them Answer Radar. Ask where the engines recommend someone else and the assistant runs the audit, reads the gaps, pulls the demand behind them, and creates the link you send a client, without you moving to another tab.

  • 37 Answer Radar tools of 65 in a single server
  • Search Console demand, audits, gaps, fixes and shares
  • Your own account and your own entitlements, not a shared key
Claude · Linkeddit connected65 tools

You ask

“Where are the answer engines recommending competitors instead of us, and what did they cite?”

gsc_queries_getreads the demand behind the questions
answer_audit_startruns them against four engines
answer_gaps_listreturns the misses, ranked
answer_share_createone link for the client
37 of the 65 are Answer Radarno copy and paste

We ship the gap, your assistant writes the fix

A tool that hands an assistant a generated paragraph is a wrapper around something the assistant already does better. What Answer Radar returns is the part it cannot get on its own: what four engines actually said, which sources they cited, and how that moved since last time.

  • The observed answer and the exact sources cited
  • Measurement that stays comparable across runs and tenants
  • Drafting stays where it belongs, in the assistant
answer_gap_get · responseevidence

Question

“Best competitor tracking tool?”

"verdict": "absent"
"recommended": ["Crayon", "Klue"]
"cited": [3 sources, with urls]
"intent": "transactional"

What comes back is the gap and the evidence. Writing the page that closes it is the assistant's job, and it is already good at that.

See the connector Included with the Compete plan

What we build, and what we refuse to

Most tools in this category ship a generator. We think that is wrapping a capability you already pay for, so we ship the measurement and let your assistant do the writing.

Data the assistant cannot get

What ChatGPT, Gemini, Perplexity and Claude actually answered on your buying questions, the sources each one cited, and the queries buyers used to reach your site. None of it is in the model.

Durability it cannot provide

Scheduled re-checks, state that survives the end of a chat, idempotent runs, and a spend position you can read before starting work.

A judgment that stays comparable

A visibility score is only useful if it means the same thing next month and for the next tenant. That is a versioned, evaluated measurement, not a model asked for an opinion.

A governed side effect

Saving a fix for approval, setting a cadence, creating a share link with a real expiry. Things that change state need to be someone's responsibility.

What we deliberately do not build

  • A copy generator. Your assistant drafts better than a template, and it can read the evidence we return.
  • A summarizer of our own output. The gap already carries the observed facts and their citation ids.
  • A chat interface on top of a dashboard. You already have one, which is the point.

The tools, grouped by the job

Thirty seven Answer Radar tools, named as they appear in the connector.

JobTools
Find the questions worth measuringgsc_queries_get, answer_prompts_propose, answer_prompts_update
Measure the enginesanswer_audit_estimate, answer_audit_start, answer_run_get
Read the missesanswer_gaps_list, answer_gap_get, answer_content_gaps_get
Close and verifyanswer_fix_draft, answer_fix_check, answer_verify_start
Track and reportanswer_metrics_get, answer_trends_get, answer_digest_get
Share the resultanswer_share_create, answer_shares_list, answer_share_revoke

The full registry, including the competitor intelligence, lead, keyword and content tools, is documented at /mcp/docs/tools.

Frequently asked questions

What is the Linkeddit MCP connector?+

It is a remote Model Context Protocol server that exposes Linkeddit's tools to an AI assistant. It carries 65 tools in one server, 37 of them Answer Radar tools for measuring and fixing AI answer visibility. It uses OAuth 2.1 with PKCE, so there is no API key to generate, store, or rotate, and it runs against your own account and entitlements rather than a shared credential.

Can I do answer engine optimization entirely inside Claude?+

Yes. Propose the buying questions from your Search Console demand, run them against ChatGPT, Gemini, Perplexity and Claude, read the gaps with the sources each engine cited, check a fix you wrote against the evidence, and create a link to send a client, all as tool calls from the assistant. The browser app and the connector are the same measurement and the same account.

Does it write the fix for me?+

No, and that is deliberate. Drafting is something an AI assistant already does well once it has the evidence, so building a generator would be wrapping a capability you already pay for. Answer Radar returns the part the assistant cannot get on its own: what the engines actually said, which sources they cited, and whether it moved after you published. Your assistant writes the page.

Which AI clients does it work with?+

Any MCP-compatible client. Claude and Claude Code, Cursor, VS Code, Windsurf, and ChatGPT are the ones we document. Setup is the connector URL plus the client id, with no client secret because it is a public PKCE client.

What does it cost?+

The Answer Radar tools require the Compete plan at $99 a month, which is the same entitlement as the browser app. There is no separate charge for using the connector.

Read next

Point your assistant at what the engines actually said

Answer Radar is included with the Compete plan, in the browser and through the connector.