agent-toolbelt-mcp

MCP server for Agent Toolbelt — gives Claude real-time stock research capabilities. Seven AI-powered analysis tools that pull live data from Polygon.io, Finnhub, and Financial Modeling Prep, then synthesize it into structured investment analysis.

Try the valuation snapshot live (no signup): www.agenttoolbelt.live

Step 1 — Get a free API key

The MCP server requires an API key. Free tier is 250 calls/month, no credit card required.

curl -X POST 'https://www.agenttoolbelt.live/api/clients/register?source=npm' \
  -H "Content-Type: application/json" \
  -d '{"email": "[email protected]"}'

The response includes your atb_... key. Save it — you'll need it in Step 2.

Step 2 — Install

Claude Code

claude mcp add agent-toolbelt \
  -e AGENT_TOOLBELT_KEY=atb_your_key_here \
  -- npx -y agent-toolbelt-mcp

Claude Desktop

Add to claude_desktop_config.json:

{
  "mcpServers": {
    "agent-toolbelt": {
      "command": "npx",
      "args": ["-y", "agent-toolbelt-mcp"],
      "env": {
        "AGENT_TOOLBELT_KEY": "atb_your_key_here"
      }
    }
  }
}

Stock Research Tools

Seven tools that give Claude live market data and AI-synthesized analysis. Each call pulls from multiple financial APIs in parallel and returns structured JSON. ~4–5 seconds per call. US-listed equities only (NYSE, NASDAQ, AMEX).

| Tool | What it does | |---|---| | stock_thesis | Full investment thesis: bullish/neutral/bearish verdict, 2–3 paragraph analysis, key strengths, risks, valuation read, insider interpretation, what to watch next earnings | | earnings_analysis | EPS beat/miss history across 5 quarters, revenue trend classification, long-term consistency read, upcoming earnings date | | insider_signal | Form 4 interpretation: distinguishes meaningful open-market purchases from routine option exercises → strong_buy to strong_sell signal + confidence rating | | valuation_snapshot | P/E, P/S, EV/EBITDA, FCF yield, ROE → cheap/fair/expensive verdict + specific buy zone | | bear_vs_bull | 3 bull + 3 bear arguments steelmanned with real data, net verdict, key debate question, who the stock suits | | compare_stocks | Head-to-head 2–3 ticker comparison: winner verdict, per-ticker strengths and concerns, recommendation map by investor goal (growth / value / quality) | | moat_analysis | Buffett-style competitive moat assessment: wide / narrow / none rating, moat sources (brand, switching costs, network effects, scale), durability outlook, and threats |

Example

Once installed, ask Claude:

"Give me a full analysis of NVDA — thesis, earnings quality, insider activity, and whether it's cheap right now."

Claude calls the tools in parallel and synthesizes a complete research note. Here's what stock_thesis returns for NVDA:

Verdict: Bullish One-liner: "Nvidia owns the essential infrastructure for the AI revolution with a defensible software moat, but the valuation demands flawless execution."

Key Strengths:

• Dominant ~80%+ data center GPU market share
• CUDA moat creates switching costs and customer lock-in
• 42 buy / 5 hold / 1 sell analyst consensus

Insider Read: Mixed — two executives bought ~47k shares each in March (positive), offset by routine selling from others.

Watch For Next Earnings: Data center revenue growth rate. Deceleration below 30% YoY would signal the boom is maturing.

Utility Tools

20 additional tools for common agent tasks.

| Tool | What it does | |---|---| | generate_schema | JSON Schema / TypeScript / Zod from a description | | extract_from_text | Extract emails, phones, URLs, dates, currencies from text | | convert_markdown | HTML ↔ Markdown conversion | | fetch_url_metadata | Title, OG tags, author, favicon from any URL | | summarize_web_page | Fetch a URL, strip boilerplate, return clean Markdown + AI summary with key points | | count_tokens | Token counts + cost estimates across 15 LLM models | | csv_to_json | CSV to typed JSON with auto delimiter and type casting | | normalize_address | US address → USPS format with component parsing | | generate_color_palette | Color palettes with WCAG scores and CSS variables | | build_cron | Natural language → cron expression with next-run preview | | build_regex | Natural language → regex with JS/Python/TS snippets | | generate_brand_kit | Full brand kit — colors, typography, CSS/Tailwind tokens | | optimize_prompt | Score and rewrite LLM prompts for clarity and specificity | | extract_meeting_action_items | Action items, decisions, and summary from meeting notes | | compare_documents | Semantic diff between two document versions | | extract_contract_clauses | Key clauses + risk flags from contracts | | mock_api_response | Realistic mock data from a JSON Schema | | audit_dependencies | CVE scan for npm/PyPI packages via OSV database | | pack_context_window | Pack content into a token budget | | strip_image_metadata | Remove EXIF/GPS metadata from images |

Pricing

• Free tier: 250 calls/month, no credit card
• Stock analysis tools: $0.02/call (PAYG)
• Most utility tools: $0.001–$0.005/call

Going to production

When the agent moves out of dev, a new set of questions shows up. What did it call last night? What arguments did it pass? Who approved the destructive one?

Cordon is an MCP gateway that sits in front of servers like this one. Point your client at Cordon instead of directly at Agent Toolbelt; Cordon forwards every call through and adds:

• A real-time audit log of every tool invocation (name, arguments, response, latency)
• Per-API-key policy: which tools each caller can use, under what conditions
• Slack-based human approvals for tool calls you've flagged as high-risk

From the agent's perspective nothing changes — same tools, same schemas. Free tier covers 250 events/month.

License

MIT