AI assistants (MCP)
Connect BrandSonar to Claude, Cursor or VS Code and ask about your mentions in plain language. Read-only: an assistant can look, never change anything.
What this is
MCP, the Model Context Protocol, is how AI assistants connect to outside tools. Once BrandSonar is connected, you can ask an assistant things like “what are people saying about us on Reddit this week?” or “how does our sentiment compare to our competitors?” and it will query your live data rather than guess.
The connector is read-only. It can search mentions, read one in full, chart sentiment over time, compare share of voice and list your alert rules. It cannot create, edit or delete anything, and it cannot see your alert destinations or any secret.
Connecting
The connector URL is the same everywhere: https://mcp.brandsonar.com/mcp. You can also copy it from AI & Agents.
Claude
Add a custom connector in Claude's settings and paste the URL. Claude opens a BrandSonar sign-in, you approve the connection, and that's it. There is no key to copy or store.
Claude Code
claude mcp add --transport http brandsonar https://mcp.brandsonar.com/mcpCursor and VS Code
Add BrandSonar to your MCP configuration:
{
"mcpServers": {
"brandsonar": {
"type": "http",
"url": "https://mcp.brandsonar.com/mcp"
}
}
}If you belong to more than one BrandSonar organisation, you'll be asked which one to connect. A connection reads from that organisation and no other. To connect a second one, add the connector again and pick it during sign-in.
What you can ask
- Recent coverage
- “Show me the most relevant mentions from the last week, and summarise the themes.”
- Sentiment shifts
- “Has sentiment about us got worse this month? What's driving it?”
- Share of voice
- “Compare our mention volume and sentiment against our competitors over 90 days.”
- Specific threads
- “Find negative mentions on Reddit mentioning pricing, and quote the strongest one in full.”
- Who to talk to
- “Which accounts write about us most often, and on which platforms?”
Access and revoking
Anyone who can sign in to your organisation can connect an assistant, and the connection only ever sees the organisation it was approved for. Admins and owners can see every connection on AI & Agents and revoke any of them.
Revoking takes effect within fifteen minutes. An assistant may complete an in-flight request in that window, after which it must reconnect.
Mentions are written by other people on public platforms, and their text goes into the assistant's context as data. A mention can contain wording designed to look like an instruction. Because the connector is read-only, the worst case is a misleading answer. Treat surprising claims the way you would treat the original post.