# buzzabout: the social media research MCP for marketers and growth engineers

Canonical page: https://buzzabout.ai/mcp

Connect buzzabout to Claude once and ask about Reddit, TikTok, YouTube, X, Instagram, LinkedIn and Facebook in plain English. Every post comes back checked for relevance, transcribed, read and labeled, with the sources cited. MCP on every plan from $98 a month, and you pay only for the posts you keep.

- Endpoint: `https://mcp.buzzabout.ai/mcp`
- Transport: streamable HTTP
- Auth: sign in with your buzzabout account (OAuth), or an `x-api-key` header
- Tools: 15, each one says in its description whether it is free or billed
- Plans: every paid plan, from Starter at $98 per month

## Two ways in: ask, or build

- **Marketers: ask Claude, get research.** Add buzzabout as a custom connector in Claude with `https://mcp.buzzabout.ai/mcp`, sign in, and ask what your audience wants, what they complain about and how they say it. You get the answer and the posts that prove it.
- **Growth engineers: vibecode on real social data.** One command: `claude mcp add --transport http buzzabout https://mcp.buzzabout.ai/mcp`. Tracking agents re-run your searches daily or weekly and deliver cited digests to a webhook or Slack, and the REST API comes with Business.

## What does this MCP server return that a scraper does not?

Most social media MCP servers hand your agent raw post records. Your agent then spends its context reading them. buzzabout does the reading on its side and returns the result.

- **Themes with post counts.** Pattern detection turns a collection into 2 to 15 named themes, focused on the question you ask. Each theme has a description, a post count and the posts behind it.
- **Answers as tables.** Ask a question about your data and get a table back: counts, breakdowns, rankings and time series.
- **Audience aggregates.** Profile the people commenting on a topic: gender, age band, interests, countries, brand affinities and Big Five (OCEAN) trait averages. No individual profile is exposed.
- **Posts already tagged.** Every collected post comes back with sentiment, emotions, topics, category, tone of voice, intent, questions, brands and a summary.

Before it collects, the server previews the topic. Collection starts only when the preview finds enough matching posts and the sample is about the right subject. Off-topic, out-of-window and duplicate posts are dropped before billing.

## What can your agent do with it?

| Prompt | Tools | Returns |
|---|---|---|
| "Collect 250 Reddit and TikTok posts about protein bars from the last three months and group the complaints." | collect_mentions, get_run, detect_patterns, query | Named complaint themes, each with a description, a post count and the posts that support it. |
| "Who comments on posts about home espresso machines? Break it down by age band and country." | collect_audience, get_run, query | A table of age bands and countries for the commenters. Groups under 10 profiles are hidden. Needs Pro or above. |
| "How did sentiment about our brand change week by week, and which posts drove the dips?" | query, list_mentions | A weekly sentiment series over posts you already collected, then the posts behind the low weeks. |
| "Research which hooks skincare creators use on TikTok this month and chart them." | ask, get_message, render | A written answer from the buzzabout research agent, plus charts and tables in hosts that support MCP Apps. Other hosts get the text. |
| "What did my tracking agents pick up since Monday?" | list_tracking_agents, query, list_mentions | The posts your scheduled tracking agents collected, with counts and themes. |

## How do I connect the buzzabout MCP server?

### Claude Code

```bash
claude mcp add --transport http buzzabout https://mcp.buzzabout.ai/mcp
```

Then run `/mcp` inside Claude Code and sign in with your buzzabout account. To use an API key instead:

```bash
claude mcp add --transport http buzzabout https://mcp.buzzabout.ai/mcp \
  --header "x-api-key: YOUR_API_KEY"
```

### Claude Desktop and claude.ai

Open Settings, then Connectors, and add a custom connector. Name it `buzzabout` and use `https://mcp.buzzabout.ai/mcp` as the URL. Click Connect and sign in with your buzzabout account.

### Cursor

`~/.cursor/mcp.json`:

```json
{
  "mcpServers": {
    "buzzabout": {
      "url": "https://mcp.buzzabout.ai/mcp",
      "headers": { "x-api-key": "YOUR_API_KEY" }
    }
  }
}
```

Clients that support MCP sign-in can leave out the `headers` block and sign in instead.

### Any streamable HTTP client

This call lists the 15 tools and costs nothing:

```bash
curl -X POST https://mcp.buzzabout.ai/mcp \
  -H "x-api-key: $BUZZABOUT_API_KEY" \
  -H "Content-Type: application/json" \
  -H "Accept: application/json, text/event-stream" \
  -d '{"jsonrpc":"2.0","id":1,"method":"tools/list"}'
```

More clients (Codex, ChatGPT, Python and TypeScript SDKs): https://docs.buzzabout.ai/mcp/use-in-your-agent

## How does authentication work?

| Method | Best for | What you need | Plans |
|---|---|---|---|
| Sign in (OAuth) | Claude Desktop, claude.ai, Claude Code, ChatGPT | Your buzzabout login | Every plan |
| API key (`x-api-key` header) | Cursor, Codex, CI jobs, your own agents | A key from Settings, then API keys, in the buzzabout app | Business and Enterprise |

The server publishes standard OAuth metadata with PKCE and dynamic client registration. A request without credentials gets a 401 that points the client to that metadata. Docs: https://docs.buzzabout.ai/mcp/auth

## Is it cheaper than building it yourself?

Yes. Research-grade social data takes five parts: a data provider for collection, speech-to-text, image and on-screen text reading, AI labeling, and hosting. At 10,000 analyzed posts a month across all seven platforms, with comments and full analysis, those parts cost about $591 at list price, before anyone builds or maintains them. The buzzabout Business plan covers the same volume for $450. (buzzabout cost model, November 2026: list-price usage plus hosting on the cheaper data provider route; engineering time not included.)

Each post can also carry up to 100 top-level comments at no extra cost, readable through the query tool.

## What does each call cost?

| Tool | What it does | Cost |
|---|---|---|
| collect_mentions | Previews a topic, then collects and tags matching posts | 1 credit per platform previewed, then 1 credit per post delivered |
| detect_patterns | Groups a collection into named themes around your question | 0.5 credit per post processed |
| collect_audience | Profiles the commenters in a collection, as aggregates | 3 credits per profile (Pro and up) |
| ingest_mention | Saves and analyses one post from its URL | 1 credit |
| query | Answers a question about your data with a table | AI inference, priced by tokens |
| ask | Runs a research turn with the buzzabout agent | AI inference, priced by tokens, plus credits for any data it collects |
| get_me | Your plan, limits and credit balance | Free |
| list_collections | Your mentions and audience collections | Free |
| list_mentions | Posts in your collections, filtered, sorted and paginated | Free |
| get_run | Progress of a collection, audience or pattern run | Free |
| get_message | Progress and output of a research turn | Free |
| render | Shows a chart or table from a research answer | Free |
| get_save_mention_status | Progress of a saved post | Free |
| list_tracking_agents | Your scheduled tracking agents and their collections | Free |
| get_tracking_agent | One tracking agent's setup | Free |

How credits work:

- You pay for posts delivered, not posts requested. Off-topic, out-of-window and duplicate posts are filtered out before billing.
- A preview is billed even when the server decides not to collect. No collection runs in that case.
- If your balance cannot cover a full run, collect_mentions collects what your balance covers instead of failing.
- Plan credits refresh every month and do not roll over. You can top up any amount from $5 at any time.
- When credits run out, new research jobs pause until you top up or your plan credits refresh.

## Which plans include the MCP server?

All of them. Prices cover the whole plan, not each seat.

| Plan | Price | Credits per month | Posts per research | Audience research | API key |
|---|---|---|---|---|---|
| Starter | $98 per month | 2,000 | 250 | No | No |
| Pro | $178 per month | 5,000 | 1,000 | Yes | No |
| Business | $450 per month | 20,000 | 1,000 | Yes | Yes |
| Enterprise | Custom | Custom | 10,000+ | Yes | Yes |

Yearly billing gives 2 months free. Try for free. Full pricing: https://buzzabout.ai/pricing (markdown: https://buzzabout.ai/pricing.md)

## What are the limits?

- Posts per run: 20 to 1,000, capped by plan: 250 on Starter, 1,000 on Pro and Business.
- Comments: up to 100 per post, 10 by default.
- Platforms: Reddit, TikTok, YouTube, X, Instagram, LinkedIn and Facebook. With no platforms set, collect_mentions searches Reddit, X and YouTube.
- History: 24 months back, or 60 months on Business.
- Collection gate: a collection starts only when the preview estimates at least 200 matching posts, returns at least 5 sample posts, and at least 15% of the sample is on topic.
- Audience data: aggregates only. Groups under 10 profiles are hidden. Individual profiles, identifiers and authenticity scores are not available over MCP.
- Waiting: get_run and get_message wait up to about 45 seconds per call for news.
- Speed: most runs finish in a few minutes. Across our production runs, a fully analysed run took a median of 3.5 minutes.

## FAQ

### What is an MCP server?

MCP (Model Context Protocol) is an open standard that lets AI assistants call outside tools. The buzzabout MCP server gives an assistant tools to collect and analyse public social media conversations.

### Is the buzzabout MCP server free?

No. It comes with every paid plan, from Starter at $98 a month. Try for free. Reads and status checks inside the server are free. Collection and analysis use credits.

### Do I need an API key?

Not for clients that sign in, such as Claude and ChatGPT. API keys are for clients you configure with a header, such as Cursor or your own agents. Keys come with the Business and Enterprise plans.

### How is this different from a social media scraping MCP?

A scraping server returns raw post records, and your agent has to read them all. buzzabout returns analysis: themes with post counts, tables that answer a question, audience aggregates and research answers. Every collected post is also tagged for sentiment, emotions, topics, brands and more.

### Can a collection run cost more than my balance?

No. If your balance cannot cover a full run, collect_mentions shrinks the run to what you can afford. If even 20 posts are out of reach, the run does not start. Paid work always needs available balance, and get_me shows your balance for free.

### What happens when a topic is too small or off topic?

The preview declines and says why: too few matching posts, too small a sample, or results about a different subject. It also says what to change. The preview is billed at 1 credit per platform, and no collection runs.

### Does it return data about individual people?

No. Audience results come back as aggregates such as age bands, countries and interests. Groups under 10 profiles are hidden, and individual profiles are never returned over MCP.

### How does the MCP server relate to the REST API?

They share the same account and data. The MCP server is on every plan. The REST API is on Business and Enterprise: https://buzzabout.ai/api

### Which AI clients work with it?

Any client that supports remote MCP servers over streamable HTTP. Setup for Claude, Claude Code, Codex, Cursor and ChatGPT: https://docs.buzzabout.ai/mcp/use-in-your-agent. Every tool: https://docs.buzzabout.ai/mcp/tools/reference