Skip to main content
Glama
mcp-octagon

mcp-octagon

Official
by OctagonAI

Octagon: MCP for Public & Prediction Markets Intelligence

Favicon The Octagon MCP server provides specialized AI-powered financial research and analysis by integrating with the Octagon Market Intelligence API, enabling users to analyze and extract insights from public filings, earnings calls, financial metrics, stocks & crypto data, stock news, and prediction markets news & research within Claude Desktop and other popular MCP clients.

Demo

Tools

✅ octagon-agent orchestrates broad market intelligence analysis

  • Public market insights: SEC filings (10-K, 10-Q, 20-F, 8-K, S-1, 13-F, DEF 14A), earnings call transcripts

  • Stocks & crypto market data: stock trading, crypto trading data, news, press release, ownerships and holdings

✅ octagon-deep-research-agent for comprehensive deep research

  • Multi-source synthesis for investment research questions

  • Best for up-to-date, cross-source thematic analysis

✅ octagon-podcasts-agent for what investors said on podcasts

  • Cited answers from 70+ finance and business shows (Bloomberg, All-In, Odd Lots, Invest Like the Best, …)

  • Filters by time window, show, host, guest, company or ticker read from the question

✅ Prediction market research tooling

  • octagon-prediction-markets-agent for updated research reports on each prediction markets

  • prediction_markets_history for structured historical market data retrieval

✅ Live Octagon documentation access

  • octagon-docs-search, octagon-docs-read, octagon-docs-list, and octagon-docs-refresh

  • MCP resources for octagon-docs://catalog, octagon-docs://status, and octagon-docs://page/{target}

Related MCP server: Signal8 MCP Server

Get Your Octagon API Key

To use Octagon MCP, you need to:

  1. Sign up for a free account at Octagon

  2. After logging in, from left menu, navigate to API Keys

  3. Generate a new API key

  4. Use this API key in your configuration as the OCTAGON_API_KEY value

Prerequisites

Before installing or running Octagon MCP, you need to have npx (which comes with Node.js and npm) installed on your system.

Mac (macOS)

  1. Install Homebrew (if you don't have it):

    /bin/bash -c "$(curl -fsSL https://raw.githubusercontent.com/Homebrew/install/HEAD/install.sh)"
  2. Install Node.js (includes npm and npx):

    brew install node

    This will install the latest version of Node.js, npm, and npx.

  3. Verify installation:

    node -v
    npm -v
    npx -v

Windows

  1. Download the Node.js installer:

  2. Run the installer and follow the prompts. This will install Node.js, npm, and npx.

  3. Verify installation: Open Command Prompt and run:

    node -v
    npm -v
    npx -v

If you see version numbers for all three, you are ready to proceed with the installation steps below.

Installation

Running on Claude Desktop

To configure Octagon MCP for Claude Desktop:

  1. Open Claude Desktop

  2. Go to Settings > Developer > Edit Config

  3. Add the following to your claude_desktop_config.json (Replace your-octagon-api-key with your Octagon API key):

{
  "mcpServers": {
    "octagon-mcp-server": {
      "command": "npx",
      "args": ["-y", "octagon-mcp@latest"],
      "env": {
        "OCTAGON_API_KEY": "YOUR_API_KEY_HERE"
      }
    }
  }
}
  1. Restart Claude for the changes to take effect

Running on Cursor

Configuring Cursor Desktop 🖥️ Note: Requires Cursor version 0.45.6+

To configure Octagon MCP in Cursor:

  1. Open Cursor Settings

  2. Go to Features > MCP Servers

  3. Click "+ Add New MCP Server"

  4. Enter the following:

    • Name: "octagon-mcp" (or your preferred name)

    • Type: "command"

    • Command: env OCTAGON_API_KEY=your-octagon-api-key npx -y octagon-mcp

If you are using Windows and are running into issues, try cmd /c "set OCTAGON_API_KEY=your-octagon-api-key && npx -y octagon-mcp"

Replace your-octagon-api-key with your Octagon API key.

After adding, refresh the MCP server list to see the new tools. The Composer Agent will automatically use Octagon MCP when appropriate, but you can explicitly request it by describing your investment research needs. Access the Composer via Command+L (Mac), select "Agent" next to the submit button, and enter your query.

Running with npx

env OCTAGON_API_KEY=your_octagon_api_key npx -y octagon-mcp

Manual Installation

npm install -g octagon-mcp

Documentation

For comprehensive documentation on using Octagon agents, please visit our official documentation at: https://octagonai.co/docs/

The documentation includes:

  • Detailed API references

  • Agent-specific query guidelines

  • Examples and use cases

  • Best practices for investment research

For the latest hosted MCP client setup guide, see:

This MCP server also exposes the live docs corpus directly to MCP clients. Documentation access is fetched from the public Octagon docs LLM entry point at https://octagonai.co/docs/llms.txt; it does not use or transmit OCTAGON_API_KEY.

Available Tools

The MCP server currently exposes the following tools:

Documentation Tools

These tools expose live Octagon docs inside the MCP session. They work even before OCTAGON_API_KEY is configured, which makes them useful for setup, troubleshooting, and discovering the right Octagon agent or API workflow.

octagon-docs-list

Lists live docs sections and pages from the Octagon docs corpus.

Parameters

  • section (string, optional): filter by docs section.

  • source (docs, site, or all, optional): defaults to the docs corpus.

  • limit (number, optional): maximum entries to return.

Searches the Octagon API, agent, MCP, and plugin docs with source URLs and optional snippets.

Parameters

  • query (string, required): search query.

  • section (string, optional): filter by docs section.

  • source (docs, site, or all, optional): defaults to the docs corpus.

  • limit (number, optional): maximum results to return.

  • includeSnippets (boolean, optional): include matched snippets in results.

Example:

Search Octagon docs for Claude plugin connector setup.

octagon-docs-read

Reads one docs page or section as Markdown by title, URL, path, or catalog id.

Parameters

  • target (string, required): docs title, URL, path, or catalog id.

  • source (docs, site, or all, optional): defaults to the docs corpus.

  • maxChars (number, optional): maximum Markdown characters to return.

  • preferCachedContent (boolean, optional): use the indexed docs corpus when available.

Example:

Read the Octagon MCP server guide.

octagon-docs-refresh

Refreshes the in-memory docs catalog from the live LLM-friendly docs endpoints.

Parameters

  • includeSite (boolean, optional): also refresh the broader https://octagonai.co/llms.txt site index.

Documentation Resources

Clients that support MCP resources can also browse:

  • octagon-docs://catalog: normalized live docs catalog.

  • octagon-docs://status: cache state, source endpoints, and refresh metadata.

  • octagon-docs://page/{target}: one docs page or section as Markdown.

octagon-agent

Orchestrates public and private market intelligence analysis.

Parameters

  • prompt (string, required): natural language research request.

  • conversation (string, optional): existing Octagon conversation ID to continue a prior octagon-agent thread. Omit this on the first turn.

  • newConversation (boolean, optional): if true, starts a fresh Octagon thread for the active session/thread anchor. Recommended for the first turn of a brand new visible chat in top-layer hosts such as Claude Desktop.

Threaded usage

octagon-agent is the only MCP tool that forwards Octagon conversation threading. It is a stateful tool and expects session continuity. The MCP resolves session/thread state in this order:

  1. stored conversation for MCP transport session identity, when the transport actually provides it

  2. stored conversation for the server-managed default stdio session

  3. explicit conversation can still override the active session conversation for that call

This package currently runs as a stdio MCP server. In stdio mode, the server automatically establishes a process-local session for continuity across calls. Most local hosts such as Claude Desktop or Cursor can therefore use octagon-agent without supplying any extra threading fields for basic follow-up behavior.

When a top-layer host knows a call is the first turn of a new visible chat, it should pass newConversation: true. That explicitly clears any stored Octagon thread for the active MCP session anchor before the call, which prevents stale continuity when a stdio host reuses the same long-lived MCP process across multiple visible chats.

This means you can use any of these patterns:

  1. First call: send only prompt

  2. Let the MCP host preserve transport session continuity or rely on the default stdio session

  3. Second call: either

    • send the new prompt in the same MCP session, or

    • keep using the same stdio MCP process, or

    • explicitly pass the previous conversation

Transport session identity is the canonical continuity primitive for standards-compliant stateful MCP transports. For local stdio usage, the server-managed process session provides default continuity.

Session identity and Octagon conversation identity are different concepts:

  • MCP session identity controls server-side continuity across tool calls

  • Octagon conversation controls the active Octagon thread inside that session

The MCP result keeps the answer in content, and also returns structured metadata for orchestrators in structuredContent:

{
  "model": "octagon-agent",
  "text": "Which stock would you like the latest price for?",
  "conversation": "conv_123",
  "responseId": "resp_123",
  "followUp": {
    "required": true,
    "inputTemplate": "<ticker or company name>",
    "instructions": "Reply with just the missing detail and reuse the conversation value from this response."
  }
}

Explicit carry-forward example:

{
  "prompt": "AAPL",
  "conversation": "conv_123"
}

New visible chat example:

{
  "prompt": "Analyze Apple",
  "newConversation": true
}

Explicit refresh example:

{
  "prompt": "Start a fresh Octagon thread for this chat",
  "newConversation": true
}

Stateful tool policy

  • octagon-agent: stateful, uses a usable continuity anchor. In stdio hosts, that defaults to the server-managed process session unless you provide explicit conversation

  • other MCP tools: stateless and may run without session continuity

Example:

Compare NVIDIA and AMD on latest quarterly revenue growth, margins, and management commentary.

More examples:

  • "What were Amazon's revenue and net income figures in Q4 2023?"

  • "Analyze Tesla's R&D spending trends over the last 3 years."

  • "What guidance did NVIDIA's CEO provide regarding AI chip demand in their latest earnings call?"

  • "Compare the price-to-earnings, price-to-sales, and EV/EBITDA ratios for the top 5 semiconductor companies."

  • "What was Anthropic's latest funding round size, valuation, and key investors?"

  • "How many investments did Andreessen Horowitz make in AI startups in the last 12 months?"

  • "What are the latest news articles for NVIDIA stock from the past week?"

octagon-deep-research-agent

Performs comprehensive multi-source deep research and synthesis.

Parameters

  • prompt (string, required): natural language research request.

Example:

Research the impact of lower interest rates on late-stage private software valuations over the next 12 months.

More examples:

  • "Extract all data fields from zillow.com/san-francisco-ca/"

  • "Research the financial impact of Apple's privacy changes on digital advertising companies' revenue and margins"

  • "Retrieve historical Bitcoin price data from 2023 and analyze the price volatility trends"

  • "Analyze the competitive dynamics in the EV charging infrastructure market"

octagon-podcasts-agent

Answers questions from what investors, strategists, operators and guests said on finance and business podcasts. Every claim cites the episode (show, title, date, link), and the answer ends with a Sources: list. Earnings calls are not podcasts: use octagon-agent for those.

Parameters

  • prompt (string, required): natural language question. Put the time window, show, host, guest, company or ticker in the question; there are no separate filter parameters.

Example:

What are podcast guests saying about NVDA since August?

More examples:

  • "What did Chamath say about AI capex on All-In recently?"

  • "Summarize the last 3 episodes of Odd Lots."

  • "Which companies were podcasts discussing most last week?"

  • "How has podcast sentiment on Fed rate cuts changed since September?"

octagon-prediction-markets-agent

Research reports on Kalshi and Polymarket prediction markets, plus market search and edge screening across both venues.

Parameters

  • prompt (string, required): a Kalshi market URL or event ticker, a Polymarket event URL or slug, or a natural-language question (which markets exist or are open for a topic, where the model disagrees with the price).

  • cache (boolean, optional): report mode for a single market.

    • omitted: octagon-prediction-markets-agent, the cached report if one exists, otherwise generate one.

    • false: octagon-prediction-markets-agent:refresh, force a fresh report (3 credits; Plus, Pro and Enterprise plans).

    • true: octagon-prediction-markets-agent:cache, read the cached report only; never generates; free. A question that names no market returns a structured JSON list of matching markets (kind: "discovery").

API errors (a missing or unresolved market reference, an ambiguous ticker, a closed market, a plan or credit limit) are returned as Error [<code>]: <message>.

Examples:

Generate a report for the Kalshi market https://kalshi.com/markets/kxbtcy/btc-price-range-eoy/kxbtcy-27jan0100
Which Kalshi NBA game-winner markets are open tonight?   (cache: true)

prediction_markets_history

Fetches historical data for a prediction market event ticker with optional pagination and time filters.

Parameters

  • event_ticker (string, required)

  • limit (number, optional)

  • cursor (string, optional)

  • captured_from (string, optional)

  • captured_to (string, optional)

  • include_analysis (boolean, optional; when true, requests analysis columns)

Example:

Fetch historical data for the Kalshi event https://kalshi.com/markets/kxbtcy/btc-price-range-eoy/kxbtcy-27jan0100

Troubleshooting

  1. API Key Issues: Ensure your Octagon API key is correctly set in the environment or config file.

  2. Connection Issues: Make sure the connectivity to the Octagon API is working properly.

  3. Rate Limiting: If you encounter rate limiting errors, reduce the frequency of your requests.

License

MIT

Individual Specialized MCP Servers

While this server provides comprehensive market intelligence combining all our specialized agents, you can also use our individual MCP servers for specific use cases:

Research Tools


⭐ Star this repo if you find it helpful!

Available Tools

3 tools
octagon-agentB

[COMPREHENSIVE MARKET INTELLIGENCE] Orchestrates all agents for comprehensive market intelligence analysis. Capabilities: Combines insights from SEC filings, earnings calls, financial metrics, stock data, institutional holdings, private company research, funding analysis, M&A transactions, investor intelligence, and debt analysis to provide holistic market intelligence. Best for: Complex research requiring multiple data sources and comprehensive analysis across public and private markets. Example queries: 'Retrieve year-over-year growth in key income-statement items for AAPL, limited to 5 records and filtered by period FY', 'Analyze the latest 10-K filing for AAPL and extract key financial metrics and risk factors', 'Retrieve the daily closing prices for AAPL over the last 30 days', 'Analyze AAPL's latest earnings call transcript and extract key insights about future guidance', 'Provide a comprehensive overview of Stripe, including its business model and key metrics', 'Retrieve the funding history for Stripe, including all rounds and investors'.

Inputs: prompt (required).

octagon-deep-research-agentB

[PUBLIC & PRIVATE MARKET INTELLIGENCE] A comprehensive agent that can utilize multiple sources for deep research analysis. Capabilities: Aggregate research across multiple data sources, synthesize information, and provide comprehensive investment research. Best for: Investment research questions requiring up-to-date aggregated information from the web. Example queries: 'Research the financial impact of Apple's privacy changes on digital advertising companies' revenue and margins', 'Analyze the competitive landscape in the cloud computing sector, focusing on AWS, Azure, and Google Cloud margin and growth trends', 'Investigate the factors driving electric vehicle adoption and their impact on battery supplier financials'.

Inputs: prompt (required).

octagon-scraper-agentB

[PUBLIC & PRIVATE MARKET INTELLIGENCE] Specialized agent for financial data extraction from investor websites. Capabilities: Extract structured financial data from investor relations websites, tables, and online financial sources. Best for: Gathering financial data from websites that don't have accessible APIs. Example queries: 'Extract all data fields from zillow.com/san-francisco-ca/', 'Extract all data fields from www.carvana.com/cars/'.

Inputs: prompt (required).

Tool Schema Changelog

Recent tool additions, removals, and schema changes observed during successful MCP inspections.

Loading tool schema changelog…

    TDQS

    B3.1/5.0

    Scored across 3 tools

    Disambiguation2/5

    The tools have overlapping purposes, causing significant ambiguity. All three tools are described as providing comprehensive market intelligence, with octagon-agent and octagon-deep-research-agent both focusing on aggregated research across multiple sources, making it unclear when to choose one over the other. The descriptions do not clearly delineate distinct boundaries, leading to potential misselection.

    Naming Consistency5/5

    The tool names follow a highly consistent pattern with the prefix 'octagon-' followed by a descriptive suffix ('agent', 'deep-research-agent', 'scraper-agent'). This uniform naming convention makes the tools easily identifiable and predictable, with no deviations in style or structure.

    Tool Count3/5

    With only 3 tools, the count feels thin for the broad scope of 'comprehensive market intelligence' covering public and private markets. While the tools aim to cover multiple data sources and analyses, the limited number may not adequately support the complex workflows implied by the descriptions, bordering on under-scoped for the domain.

    Completeness2/5

    There are significant gaps in the tool surface for market intelligence. The tools focus on aggregation and scraping but lack dedicated operations for specific actions like updating data, deleting records, or managing user queries, which are essential for a complete CRUD lifecycle. This incompleteness could lead to agent failures in handling varied tasks.

    Maintenance

    ActivityMaintained
    ResponsivenessNo recent community issues

    Related MCP Connectors

    Related MCP Servers

    • F
      license
      Not graded
      quality
      Not graded
      maintenance
      Provides access to a comprehensive financial intelligence platform featuring real-time market data, quantitative models, and alternative data sources. It enables users to perform advanced financial analysis including options analytics, portfolio modeling, and SEC filing research.
      -
    • A
      license
      Not graded
      quality
      B
      maintenance
      Financial data and research MCP for AI agents: filings with full-text and fact search, statements as reported, earnings, insider and institutional ownership, corporate events, executives, analyst data, company discovery and research signals for US, China and Japan equities. Every figure traced to its filing. Browser sign-in.
      10
      MIT