Octagon MCP Server

repository·main·Indexed 40 days ago

https://github.com/octagonai/octagon-mcp-server

An MCP (Model Context Protocol) server for the Octagon API that provides AI-powered financial research for public and private markets. It enables LLM clients like Claude Desktop and Cursor to access SEC filings, earnings calls, crypto/stock data, and prediction market research via specialized tools including the octagon-agent, octagon-deep-research-agent, and octagon-prediction-markets-agent.

Tokens
10.7K
Snippets
31
Records
59
Agent score
50%

What's inside Octagon

  1. Overview of Octagon MCP Tools

    main

    The Octagon MCP server provides AI-powered financial research and analysis by integrating with the Octagon Market Intelligence API. It exposes several specialized agents and tools to MCP clients:

    • octagon-agent: Orchestrates broad market intelligence, including SEC filings (10-K, 10-Q, 20-F, 8-K, S-1, 13-F, DEF 14A), earnings call transcripts, and stocks/crypto market data (trading data, news, press releases, ownerships).
    • octagon-deep-research-agent: Performs comprehensive multi-source synthesis for investment research and thematic analysis.
    • octagon-prediction-markets-agent: Provides updated research reports on prediction markets.
    • prediction_markets_history: Retrieves structured historical market data.
    • Documentation Tools: octagon-docs-search, octagon-docs-read, octagon-docs-list, and octagon-docs-refresh allow direct access to Octagon documentation via MCP resources like octagon-docs://catalog, octagon-docs://status, and octagon-docs://page/{target}.
  2. Route investment research requests to specific Octagon skills

    main

    When using the octagon-analyst-master workflow, requests should be routed to the following specialized skills based on the specific data needed:

    • Estimates and consensus expectations: analyst-estimates
    • Kalshi market reports or event history: prediction-markets-analysis
    • Earnings transcript and guidance analysis: earnings-call-analysis
    • Annual filing and risk review: sec-10k-analysis
    • Real-time pricing and market snapshot: stock-quote

    If a request remains broad after attempting to route to these specific tools, use the octagon-agent tool.

  3. Analyze analyst estimate outputs

    main

    When the analyst-estimates tool returns data, it provides a table of future periods containing revenue and EPS ranges, averages, and analyst coverage. It also includes a summary of growth trajectory, estimate dispersion, and coverage quality.

    Use these heuristics to interpret the data:

    • Implied growth: Compare future estimates against current actual results.
    • Dispersion: Wider ranges in estimates imply lower analyst conviction.
    • Coverage depth: A higher number of analysts typically indicates a stronger consensus.
    • Time horizon: Confidence in estimates should be treated as lower for periods further in the future.
  4. Expected output format for SEC 10-K analysis

    main

    When using the sec-10k-analysis skill, the output is structured to provide high-density intelligence rather than exhaustive lists. Expect the following structure:

    • Business Overview: A concise summary of the company's operations.
    • Material Risks: A curated list of the most significant risks (not an exhaustive list of all risks).
    • Financial Details: Highlights of reporting segments, capital allocation strategies, and accounting watch items.
    • Comparative Analysis: Comparison of important changes to the prior annual filing (where evidence is available).
  5. Expected output for earnings call analysis

    main

    When using the earnings call analysis skill, the output is structured to provide the following:

    • Segmentation: Separation of prepared remarks from the Q&A session where possible.
    • Guidance: Clear quotes or paraphrases of management guidance.
    • Thematic Analysis: Identification of repeat themes, risk signals, and changes compared to prior calls.
    • Monitoring: A concluding summary of the top issues to monitor in the upcoming quarter.
  6. Use Octagon documentation tools for setup and discovery

    main

    Before starting analysis, use the following documentation tools to answer questions about setup, capabilities, or how Octagon works. Always prefer documentation context over model memory for questions regarding installation, connectors, authentication, and tool surfaces.

    • Discovery: Use octagon-docs-search to find information on topics like Claude plugin, MCP server, available agents, or authentication.
    • Canonical Context: Use octagon-docs-read to retrieve detailed information from specific documents such as Octagon Claude Plugin, Octagon MCP Server, Octagon Agents Guide, or How Octagon API Works.
  7. Use Octagon Analyst Master for broad investment research

    main

    The octagon-analyst-master skill is designed to handle broad investment research requests that span multiple specialized analyst workflows. Use this skill when a user request requires a full company brief, an investment memo, or a mix of data types including filings, estimates, earnings, valuation, and market data. It acts as a router to determine which specific Octagon skill is best suited for the task.

    {
      "server": "octagon-mcp",
      "toolName": "octagon-agent",
      "arguments": {
        "prompt": "Build an investment analyst brief on NVDA covering business quality, latest performance, guidance, valuation context, and key near-term risks."
      }
    }
  8. Configure Octagon MCP for Claude Desktop

    main

    To use Octagon in Claude Desktop, add the server to your claude_desktop_config.json file. Replace YOUR_API_KEY_HERE with your actual Octagon API key.

    1. Open Claude Desktop.
    2. Navigate to Settings > Developer > Edit Config.
    3. Add the following configuration:
    {
      "mcpServers": {
        "octagon-mcp-server": {
          "command": "npx",
          "args": ["-y", "octagon-mcp@latest"],
          "env": {
            "OCTAGON_API_KEY": "YOUR_API_KEY_HERE"
          }
        }
      }
    }
    1. Restart Claude Desktop.
    {
      "mcpServers": {
        "octagon-mcp-server": {
          "command": "npx",
          "args": ["-y", "octagon-mcp@latest"],
          "env": {
            "OCTAGON_API_KEY": "YOUR_API_KEY_HERE"
          }
        }
      }
    }
  9. Run Octagon MCP via npx or Manual Installation

    main

    Run with npx

    You can run the server directly using npx by providing the API key in the environment:

    env OCTAGON_API_KEY=your_octagon_api_key npx -y octagon-mcp

    Manual Installation

    To install the package globally via npm:

    npm install -g octagon-mcp
    env OCTAGON_API_KEY=your_octagon_api_key npx -y octagon-mcp
  10. Use the analyst-estimates skill to retrieve financial projections

    main

    The analyst-estimates skill allows you to retrieve analyst revenue and EPS (Earnings Per Share) expectations for a public company. This is useful for analyzing forward expectations, consensus assumptions, or valuation inputs.

    When using this skill, you should specify the <TICKER>, the period (e.g., annual), and constraints like the number of records and page number to ensure precise data retrieval.

    {
      "server": "octagon-mcp",
      "toolName": "octagon-agent",
      "arguments": {
        "prompt": "Retrieve analyst financial estimates for AAPL for the annual period, limited to 10 records on page 0."
      }
    }
  11. Configure Octagon MCP for Cursor

    main

    Requires Cursor version 0.45.6 or higher.

    1. Open Cursor Settings.
    2. Go to Features > MCP Servers.
    3. Click + Add New MCP Server.
    4. Enter the following details:
      • Name: octagon-mcp
      • Type: command
      • Command: env OCTAGON_API_KEY=your-octagon-api-key npx -y octagon-mcp (Replace your-octagon-api-key with your actual key).

    Windows Users: If the above command fails, use: cmd /c "set OCTAGON_API_KEY=your-octagon-api-key && npx -y octagon-mcp"

    After adding, refresh the MCP server list. You can use the tools via the Composer Agent (Command+L on Mac) by selecting "Agent" next to the submit button.

    env OCTAGON_API_KEY=your-octagon-api-key npx -y octagon-mcp