Blog/ Connect Your AI Coding Assistant to AskEdgar with MCP

Connect Your AI Coding Assistant to AskEdgar with MCP

JJason TangeOct 19, 2026

AskEdgar's API docs include a public MCP server, so an AI coding assistant can look up endpoints, read their parameters and response fields, and help you write working code against them. You don't need an API key to explore the docs this way. You only need one when you start requesting data.

This guide covers what the MCP server does, how to connect it, and how to go from exploring endpoints to a working research tool.

What you need

  • An AI tool that supports remote MCP servers. The AskEdgar docs name Codex, which supports remote HTTP MCP connections and custom headers. Other clients work too if they support remote MCP, but connection options vary by client.
  • The server URL: https://askedgar.readme.io/mcp
  • For data requests only: an AskEdgar account, an API key from the API platform and credits on your account. Exploring endpoints and reading docs needs no key.

Connect the server

  1. Add the URL. In your AI tool's remote MCP settings, add https://askedgar.readme.io/mcp. Check your tool's own MCP documentation for where that setting lives. For Codex, see OpenAI's Codex MCP docs.
  2. Check the tools loaded. Your assistant should now have four AskEdgar tools:
  • list-endpoints (no key needed): lists API paths and methods.
  • search-endpoints (no key needed): finds endpoints and parameters by keyword.
  • get-endpoint (no key needed): shows an endpoint's parameters, response schema and auth requirements.
  • execute-request (key needed for data requests): runs an API request.
  1. Add your key, if your tool supports it. AskEdgar data endpoints authenticate with an API-KEY header. ReadMe's generic examples show Authorization: Bearer, which is a different scheme, so use API-KEY for AskEdgar data requests. Header options vary by client. Before you build anything recurring, run one paid request and confirm it works.

If your tool can't send the header, you can still use MCP for discovery. Have your assistant read the endpoints, then write server-side code that calls the REST API directly, as in the Quick Start.

Try it: three prompts

Start with discovery, which is free, and only move to data requests once you know what you need. These are starting points, so results will vary by tool and model.

1. Explore (no key needed)

Use the AskEdgar MCP tools to find the float, dilution and registrations endpoints. Summarize each one's parameters and key response fields, then propose a simple small-cap research dashboard that uses them.

2. Check cost before building

Using the AskEdgar free usage estimate endpoint, estimate the cost of pulling dilution data for a watchlist of 20 tickers once a day. Show me the estimate before writing any code that requests data.

3. Build from a starter

Read the README of the AskEdgar public dashboard starter linked from the AskEdgar docs (github.com/jasontange/askedgar-dilution-mobile). Set it up so my API key is stored in server-side secret settings, test it with one ticker, then help me customize the watchlist.

Tip from the docs: on dilution endpoints, show_in_play=true can significantly reduce response cost.

Good to know

  • Credits: data requests through MCP use your API credits and endpoint permissions, the same as direct API calls. Reading docs and discovering endpoints is free.
  • Keep your key out of prompts. Store it in server-side secret settings, never in a prompt, a public repo or browser-side code.
  • Access varies by endpoint. Some endpoints have their own filter and access rules. Ask your assistant to check an endpoint's requirements with get-endpoint before you build on it.
  • Machine-readable docs: an index of the docs is also available at askedgar.readme.io/llms.txt.

Common questions

What is MCP? The Model Context Protocol is an open standard that lets AI assistants connect to outside tools and data. AskEdgar's MCP server gives your assistant tools to explore and call the AskEdgar API.

Do I need an API key? Not to explore endpoints, read the docs or estimate a query's cost. Those are free. You need a key and credits to request data. Without a key, data endpoints return a missing_api_key error.

Does using MCP cost credits? Discovery is free. Data requests through MCP use credits the same way direct API calls do.

Which AI tools work? Any tool that supports remote MCP servers can connect for discovery. To request data, your tool also needs to send a custom API-KEY header. The AskEdgar docs name Codex as supporting this. For other tools, check their MCP documentation.

What if my tool can't send the header? Use MCP to explore, then have your assistant write server-side code that calls the REST API directly.

Why does the URL show an error in my browser? Opening it in a browser shows "This URL can only be accessed with a MCP client." That's expected. Add the URL in your AI tool's MCP settings instead.

Last checked against the AskEdgar docs on October 19, 2026. MCP support changes quickly, so the MCP / AI guide has the latest details.

Start building

Connect the MCP server and explore for free. When you're ready to pull data, get your API key and read the MCP / AI guide for the latest setup notes.