Peak Answer is an MCP server as well as an app. Ask your assistant how your visibility moved this week, which questions name a competitor, or to track a new one — without opening the dashboard.
Most products that advertise an AI integration mean a chatbot bolted onto their own interface. This is the other direction. The connector exposes the product’s own tools to whichever assistant you already work in, so the question “are we losing any buying questions this week?” gets answered in the place you were already typing, by an assistant that can then go and do something about it.
It is a remote server over streamable HTTP. There is nothing to install and nothing to run locally; a client that speaks MCP connects to a URL and authenticates.
Nine tools. Six of them read your own measured data, and three work on any public URL, which means they are useful before you have any history at all.
my_brand | Which domain this connection is scoped to. |
my_visibility | Your visibility index and how it has moved. |
my_questions | Every question being tracked, and which ones you lose. |
my_question_history | One question over time, run by run. |
my_actions | The work the product is recommending, in order. |
my_search_console | Traditional search performance alongside the AI measurement. |
audit_site | Answer readiness, crawler access, robots, sitemap, schema, headings, readability. |
geo_audit | Ten real buying questions asked of a real model. |
generate | llms.txt, FAQ schema, a content brief, a prompt set or semantic keywords. |
Eight more. A connector that can only read turns the assistant into a reporting surface, which is half a product: you find out a question is being lost and then have to go somewhere else to act on it. These close that loop.
track_question | Start measuring a question. |
untrack_question | Stop measuring one. |
mark_action_done | Close an action. |
skip_action | Decline one, with the reason recorded. |
add_competitor | Add a brand to the competitive set. |
remove_competitor | Take one out. |
save_context | Tell the product something about your business that it should use. |
approve_article | Approve a draft for publishing. |
Every tool carries annotations, so a client knows which are read-only, which are idempotent, and which change something outside the system. approve_article is the only one marked destructive, because approving a draft for publishing is a decision that should prompt a confirmation.
The useful shape is a question that would take four clicks in a dashboard:
One line in Claude Code, or a URL in Claude Desktop and ChatGPT’s connector settings. The step-by-step, per client, with the authentication options and the troubleshooting, is in the guide:
There is also reference documentation covering the transport, the auth flow and every tool’s parameters, which is the page to send a developer to.
Published in the official MCP registry as io.github.peakanswer/peak-answer, and listed on Glama with ownership verified. Both mean the same thing in practice: the server is discoverable by clients that browse registries rather than being configured by hand.
The Model Context Protocol, an open standard for connecting tools to AI assistants. A client that speaks it — Claude, ChatGPT, Cursor and a growing list of others — can call a server's tools during a conversation.
No. The connector is included on every plan, including the trial.
Both. Nine read tools and eight write tools. The write tools cover tracking questions, managing competitors and actions, saving context and approving articles, which is most of what the app itself does.
approve_article is the one tool marked destructive and open-world in its annotations, so a well-behaved client will confirm before calling it. Approval means a draft is cleared for publishing, which is a decision worth a confirmation step.
OAuth 2.0, with dynamic client registration and client ID metadata documents both supported, or an API key from Settings. Each connection is scoped to exactly one brand.
Including the trial. Connect it on day one and ask it what it found.
Start a 7 day trial