@getbeta/mcp

MCP server for BETA — cited company intelligence for AI agents.

What is BETA?

BETA is a company-intelligence API where every claim is corroborated and cited. It covers 673 curated companies (Fortune 500 + tech leaders) and ~7,200 SEC-filer entities, with news + filings + people pipelines feeding fresh signals. Built for systems where being wrong has cost.

What this MCP gives you

Four tools an LLM can call:

| Tool | Purpose | |------|---------| | beta_validate_company | Verify a company exists in BETA's dataset; get canonical key + confidence | | beta_enrich_company | Firmographics, market cap, SWOT, competitors, events, people | | beta_get_company_timeline | Material events (M&A, leadership, IPO, restatements) over a date range | | beta_get_company_people | Indexed executives, board, founders, investors |

Coverage

• 673 curated companies — Fortune 500 + selected private + tech leaders
• ~7,200 SEC-filer entities — automatically tracked from SEC filings
• Fresh signals — RSS pipelines on 186 sources; SEC 8-K monitoring; people-mention indexing
• Citations on every response — every claim links to its source

Install

npx -y @getbeta/mcp

Or install globally:

npm install -g @getbeta/mcp

Configure (Claude Desktop)

Add to ~/Library/Application Support/Claude/claude_desktop_config.json (macOS) or %APPDATA%\Claude\claude_desktop_config.json (Windows):

{
  "mcpServers": {
    "beta": {
      "command": "npx",
      "args": ["-y", "@getbeta/mcp"],
      "env": {
        "BETA_API_KEY": "your-api-key-here"
      }
    }
  }
}

Configure (Cline / Continue / Cursor / other MCP clients)

The server speaks stdio. Any MCP-compatible client can launch it:

command: npx
args: ["-y", "@getbeta/mcp"]
env:
  BETA_API_KEY: your-api-key-here

Or directly:

BETA_API_KEY=... node /path/to/@getbeta/mcp/dist/index.js

The MCP talks to https://api.getbeta.io by default. Set BETA_API_URL if you need to point at staging or a self-hosted instance.

Get an API key

Sign up at getbeta.io. The Free tier includes 100 calls/month — enough to evaluate the MCP and integrate it into a prototype. Paid tiers start at $19/month.

Example agent loop

A typical Claude Desktop interaction using these tools:

You: I have a meeting with Nvidia tomorrow about partnership opportunities. Give me a quick brief on their current strategic priorities and any recent material events.

Claude (internally):

1. Call beta_validate_company("Nvidia") → confirms canonical entity, returns key
2. Call beta_enrich_company("Nvidia", include=["intelligence", "events"]) → returns context, events with sources
3. Synthesize a brief, with citations, so you can verify any claim before walking into the meeting

The MCP makes the cited part automatic. Every fact comes with a source URL.

What you get back

A trimmed beta_enrich_company("Nvidia") response (real shape — see docs for the full schema):

{
  "verified": true,
  "confidence": 1,
  "entity": {
    "key": "nvidia",
    "name": "NVIDIA Corporation",
    "type": "public_company",
    "ticker": "NVDA",
    "industry": "technology",
    "publicUrl": "https://getbeta.io/company/nvidia"
  },
  "scale": {},
  "context": "NVIDIA's CEO Jensen Huang outlined a $1 trillion revenue opportunity in artificial intelligence chips through 2027 during the company's GTC conference, signaling confidence in sustained demand across enterprise and consumer applications [2][6]. The presentation generated signific…",
  "events": [],
  "people": [],
  "meta": {
    "enrichedAt": "2026-05-19T23:01:01.650Z",
    "sources": ["entity_resolution", "entity_summaries"],
    "resolution": { "source": "curated", "confidence": 1 },
    "costEstimate": 0.008
  },
  "headlineInsight": {
    "insight": "NVIDIA is pivoting from pure compute dominance into an integrated AI infrastructure stack—combining chips (Vera Rubin platform), software (NemoClaw agents, DLSS 5), and edge deployment (DGX Station, B…",
    "confidence": "HIGH",
    "temporal_grounding": "GTC keynote and product announcements, March 2026",
    "supporting_articles": [1, 3, 6, 7, 8, 12],
    "meta": {
      "entity": "nvidia",
      "article_count": 265,
      "generated_at": "2026-03-17T17:43:27.913Z"
    }
  },
  "intelligence": null
}

Notice the [N] markers inside context and the supporting_articles array in headlineInsight: every claim resolves to an indexed source article. meta.sources and meta.resolution tell you how the entity was matched and which subsystems produced the response. events, people, and intelligence populate when the underlying signals exist — empty arrays here mean BETA didn't have indexed events/people for Nvidia at request time, not that it made any up.

Need help?

• Installing or configuring → check this README, then developers.getbeta.io once it's live
• Possible bug → file a bug
• Want a new tool / feature → file a feature request
• Paid customer, urgent → [email protected] (faster response)
• Security concern → see SECURITY.md

Status & roadmap

See CHANGELOG.md for release history. The next planned release (v0.2.0) adds an examples/ folder, citation-shape docs in the README, and fresh-shell install verification.

License

MIT