Skip to main content
Reputably tracks what people say about a business: Google reviews, brand mentions across X, Reddit, YouTube, Facebook communities and the open web, the sales leads hiding in those mentions, and whether AI assistants recommend the business when someone asks. All of it is reachable two ways, and both land on the same permissions.

Connect an assistant

Point Claude, Cursor or any MCP client at Reputably and ask questions in plain language. Nothing to build.

Call the API

Create an agent key and call the HTTP endpoints from your own scripts, CI or backend.

​
What you can do

  • Read reviews across every connected location, including which ones still need a reply.
  • Read brand mentions from every platform in one feed, each with sentiment and relevance already scored.
  • Read leads with the public post URL, so a human can go and reply where the conversation is.
  • Read AI visibility: which assistants mention the brand, what they cite, and how that moves.
  • Generate drafts: review replies, mention replies, post drafts and insight summaries.
  • Queue reports and download finished ones as CSV.

​
What you cannot do

The surface is deliberately read plus safe writes. No credential can send anything to an external platform, trigger a sync, touch billing or manage other credentials. This is enforced on the server, not in the client, so it holds however you connect. See What agents cannot do.

​
Start here

1

Create a credential

Sign in, then open Settings → API & MCP. Claude clients can sign in instead, with no key to copy. See Authentication.
2

Connect or call

Paste the connection snippet for your client, or call the HTTP API with a bearer token.
3

Find the right workspace

An agency keeps each client in its own workspace, and every read is scoped to one of them. Workspaces explains the trap that empties your results.