> ## Documentation Index
> Fetch the complete documentation index at: https://docs.chicago.global/llms.txt
> Use this file to discover all available pages before exploring further.

# Parallax MCP Server

> Use Parallax investment intelligence directly from Claude, Cursor, VS Code, and other AI clients

The **Parallax MCP server** exposes Chicago Global's investment research as a set of
tools any [Model Context Protocol](https://modelcontextprotocol.io) client can call —
Claude, Cursor, VS Code, ChatGPT, and more. Instead of wiring up the
[REST API](/api-reference/introduction) yourself, you connect once and your AI assistant
can score stocks, analyze portfolios, pull financials, read macro and market telemetry,
synthesize news, and generate research reports on your behalf.

<CardGroup cols={2}>
  <Card title="Connect a client" icon="plug" href="/mcp/connect">
    Add Parallax to Claude, Cursor, VS Code, or ChatGPT in a couple of minutes.
  </Card>

  <Card title="Authentication" icon="lock" href="/mcp/authentication">
    How OAuth sign-in and programmatic bearer tokens work.
  </Card>

  <Card title="Tool reference" icon="toolbox" href="/mcp/tools">
    Every tool the server exposes, grouped by what it does.
  </Card>

  <Card title="REST API" icon="code" href="/api-reference/introduction">
    Prefer raw HTTP? The same intelligence is available over the REST API.
  </Card>
</CardGroup>

## Endpoint

The server speaks **Streamable HTTP** (the current MCP transport). Point your client at:

```
https://mcp.chicago.global/api/mcp
```

A legacy **SSE** endpoint is also available for older clients:

```
https://mcp.chicago.global/api/sse
```

<Note>
  When connecting with an OAuth-capable client, sign in through your browser
  to link your Parallax account. The client then manages the access token —
  see [Authentication](/mcp/authentication).
</Note>

## What you can do

| Area                   | Examples                                                                                   |
| ---------------------- | ------------------------------------------------------------------------------------------ |
| **Stock intelligence** | Score a stock vs. its peers, historical factor trends, company profile, technical analysis |
| **Portfolio**          | Quick factor scores, full multi-stock analysis, redundancy/overlap checks                  |
| **Fundamentals**       | Income statement, balance sheet, cash flow, ratios, valuation summary                      |
| **ETFs**               | Profile & factor scores, top holdings look-through, search, daily prices                   |
| **Macro & markets**    | Country macro outlook, daily market telemetry, leading/lagging baskets                     |
| **News & research**    | AI-synthesized stock news, impact assessments, full research reports                       |
| **Discovery**          | Build a thematic stock universe from a plain-English theme, search stocks & ETFs           |
| **Docs**               | Fetch Parallax methodology and glossary pages inline                                       |

See the full [tool reference](/mcp/tools) for the complete list, including which tools
are free vs. paid and which run asynchronously.

## Symbol format

Tools accept symbols in **RIC** format (ticker + exchange suffix, e.g. `AAPL.O`,
`JPM.N`, `2330.TW`). Many tools also accept a plain ticker (`AAPL`) and resolve it for
you; when in doubt, use the `search_stocks` tool to find the exact symbol first.

## Relationship to the REST API

The MCP server is a thin, auth-aware layer over the same
[`api.chicago.global`](/api-reference/introduction) backend. Tool outputs mirror the REST
responses. Long-running tools normally poll their background jobs internally and return
the completed result. Use `check_job_status` only if a tool returns a job ID without a
completed result. If you're building an application, use the REST API directly; if
you're working inside an AI client, use MCP.