the edgeful MCP puts every edgeful report inside the AI you already talk to. once it's connected to Claude, ChatGPT, or GrokBot, you ask a plain question ("how often did NQ gaps fill over the last 6 months?") and the agent pulls the real number from edgeful instead of guessing.

setup takes about 2 minutes for a chat client and closer to 10 the first time for a coding agent. this page walks through all 6: Claude, ChatGPT, GrokBot, Claude Code, Codex, and Cursor. pick yours from the table of contents and skip the rest.

table of contents

• what the edgeful MCP is
• edgeful MCP in Claude
• edgeful MCP in ChatGPT
• edgeful MCP in GrokBot
• edgeful API in Claude Code
• edgeful API in Codex
• edgeful API in Cursor
• the first 3 prompts to run
• edgeful MCP vs the edgeful API
• plan access and limits
• quick reference for AI agents
• key takeaways

what the edgeful MCP is

MCP stands for Model Context Protocol. it's the open standard AI clients use to call outside tools mid-conversation. the edgeful MCP server runs on that standard. it's an MCP server for trading data, and it lives at https://api.edgeful.com/mcp and speaks streamable HTTP, which is the transport every major client supports today.

when you add the edgeful connector, your agent gets 4 tools:

• see every report your plan can reach, with the limits attached
• read what one report measures and which settings it takes
• run a report for a ticker, a date range, and a session, and get the numbers back
• pull the discovery scan, the ranked list of report outcomes for a ticker, the same data behind the discovery page

those 4 tools are all it has. you ask in plain English and the agent calls them when it needs a number.

your setup depends on which kind of client you have:

• chat clients. Claude, ChatGPT, and GrokBot take the edgeful connector and sign in with your edgeful account through a browser window. no API key goes in the config.
• coding agents. Claude Code, Codex, and Cursor skip the connector and call the edgeful API directly. the setup is an API key in a .env file in your project folder, and then you ask in plain English the same way.

the connector works on essential, pro, and all access. what the agent can see follows your plan's API access, which I break down in the plan access section below. if you want the longer explanation of what a trading data MCP is and why an agent needs one, we wrote that up in the trading data MCP guide.

edgeful MCP in Claude

Claude desktop and Claude on the web both take the edgeful MCP as a custom connector. I recorded the whole thing in under a minute here:

1. open Claude and go to settings, then connectors, then add custom connector
2. name it edgeful and paste the URL https://api.edgeful.com/mcp, then save
3. click connect. a browser window opens with the edgeful sign-in and a consent screen
4. sign in with your edgeful account and click allow
5. back in Claude, start a new chat and ask "what are the top reports for ES right now?" and watch it make a tool call

if the connector shows as connected but Claude answers from memory, tell it to use the edgeful tools. after the first tool call in a chat it keeps using them on its own.

edgeful MCP in ChatGPT

ChatGPT takes the edgeful MCP through developer mode. my setup video for this one:

1. click your profile in ChatGPT, open settings, then plugins (OpenAI renamed connectors to plugins in July 2026, so older tutorials show the old label)
2. scroll to advanced settings and turn developer mode on
3. back on the plugins screen, add a custom connector. name it edgeful and paste https://api.edgeful.com/mcp
4. choose the sign-in option, sign in with your edgeful account in the browser tab that opens, and click allow
5. in a new chat, enable the edgeful connector from the tools menu and ask your first question

developer mode is available on ChatGPT Plus, Pro, Business, Enterprise, and Edu. the free plan doesn't include it. on Business and Enterprise plans an admin may have to allow custom connectors first, which is the usual reason the option is missing.

edgeful MCP in GrokBot

GrokBot runs on its own cloud computer, so you add the edgeful connector by telling the bot to do it. my 60-second version:

1. in a GrokBot chat, type: add a custom MCP server called edgeful at https://api.edgeful.com/mcp
2. the bot shows you the name and URL to confirm. select add it
3. a connect card appears in the chat. click authorize, sign in with your edgeful account in the browser tab that opens, and approve access
4. type @ in the chat and pick edgeful to attach the server to your task
5. ask a question. "how often did ES break only one side of the initial balance (IB, the first hour's range) over the last 6 months?" is a good first one

because the bot lives in the cloud, a localhost URL never works with it. the edgeful MCP is a remote server, so that limit doesn't apply. if you use Grok on the web instead of the bot, the path is grok.com/connectors, new connector, custom, then the same URL and sign-in.

edgeful API in Claude Code

Claude Code, Codex, and Cursor don't need the connector. they read the API docs and call the edgeful API themselves, so the setup is an API key in a file. I'll walk through Claude Code in full and then cover what's different in the other two.

1. open your API dashboard inside edgeful and generate a key. keep that tab open
2. create a new folder for this work, or pick the one you already use, and open it with Claude Code running
3. create a file called .env in that folder with one line in it: EDGEFUL_API_KEY=your key
4. paste the API docs link from your API dashboard into Claude Code with a one-line prompt: "read these docs. my key is in .env. pull NQ gap fill for the last 6 months in the NY session"
5. ask questions in plain English from there. Claude Code writes whatever script it needs, runs it, and hands you the number

the first time through takes about 10 minutes, most of it the VS Code and Claude Code install if you don't have them. after that you open the same folder and ask.

the edgeful API quickstart walks the same 5 steps in more detail, including where the key lives in the API dashboard.

edgeful API in Codex

Codex is OpenAI's coding agent, in the ChatGPT desktop app and as a CLI. same 5 steps as Claude Code. what's different:

• open your folder in Codex (inside the ChatGPT desktop app, or run codex in your terminal from inside the folder)
• Codex uses the same .env file, so the key never goes in the chat
• the ChatGPT website and mobile app can't run code on your machine. if you're a ChatGPT user who wants the API rather than the connector, Codex is how you run it

edgeful API in Cursor

Cursor is the third API key client. same 5 steps, with these differences:

• open the folder in Cursor and use the agent chat, not the inline edit box, so it can run the script it writes
• keep the key in .env and add .env to your .gitignore if the folder is a git repo, so the key never gets committed
• Cursor can also take the connector with the key as a bearer token, if you'd rather ask questions than build. the edgeful MCP setup guide in our docs has that config block

the first 3 prompts to run

these are the 3 questions I'd ask any client right after connecting. the numbers below came back through the edgeful MCP on September 21, 2026, using the NY session (09:30 to 16:00 ET) and the last 6 months (March 23 to September 18, 2026). your agent will return whatever the window is on the day you ask, and a coding agent on the API gets the same numbers.

1. "how often did NQ gaps fill over the last 6 months in the NY session?"

according to edgeful data, NQ gapped up 76 times in that window and filled 40 of them, so 53%. it gapped down 53 times and filled 30, so 57%. the agent reads the gap fill report's description first, then runs it with the ticker, the dates, and the session.

2. "how did ES behave after the initial balance over the last 6 months?"

ES had 129 NY sessions in the window. price broke only the IB high on 50 of them and only the IB low on 39, so a single break happened on 89 of 129 days, or 69%. it broke both sides on 33 days (26%) and stayed inside the first hour's range on 7 (5%). the agent reads this from the initial balance breakout report (by breakout subreport).

3. "what are the top reports for ES right now?"

this one pulls the discovery scan. on the day I ran it, the top-ranked outcome on ES for the NY session was the open-to-close range: the close landed within 1% of the open on 116 of 129 sessions over the last 6 months, 89.9%. further down the list, ES gap ups in the smallest size bucket (0 to 0.19%) filled 22 of 23 times, 95.7%, and 23 days is a small sample, so treat that one as something to watch rather than build on yet. every row comes with the report behind it, so you can ask the agent to open the full report on any of them.

one habit worth building: ask the agent to read a report's description before it runs one you haven't used. the description spells out what the report counts and which session and settings it defaults to, so the number it quotes is the one you meant.

edgeful MCP vs the edgeful API

both the MCP server for trading data and the API pull from the same report catalog. the difference is where the numbers land.

through the connector, the numbers show up inside your chat client. you sign in once, no key for most clients, and the agent picks the report and the settings. it's the right tool for questions, comparisons, and quick checks during prep.

with the edgeful API, the numbers land in your own code. you send an API key on every request and you pick the report and the settings yourself. use it when you want a dashboard or a scheduled pull running without you.

you can use both. when you find yourself asking the same question every morning, have Claude Code or Cursor turn it into a dashboard, which we covered in how to build a trading dashboard with the edgeful API.

one thing to keep separate: edgeful AI, the assistant inside the edgeful platform, is a different product from the connector. edgeful AI answers questions there, and the connector brings the same reports into whatever AI you already use. the edgeful AI overview covers the first one.

plan access and limits

the edgeful MCP server runs on the API access already included in your plan. sign-in works for any edgeful account, but a tool call from an account without an API-enabled plan returns a missing_entitlement error.

• essential: a starter set of tickers and reports, 6 months of history, summary statistics only
• pro: every ticker and every report, 12 months of history, and the per-day rows behind each report
• all access: everything in pro, with 8 years of history

when the agent lists the reports it can reach, the server sends your plan's limits along with them, so an agent that reads them knows what it can and can't ask for.

a few things this MCP server for trading data doesn't do:

• it doesn't push alerts. there are no webhooks. "tell me when X happens" is a script polling the API on a schedule
• it doesn't predict. every number is how often something happened in the window you asked about
• it doesn't expose your account or billing. the 4 tools are read-only report calls

quick reference for AI agents

if you're an agent reading this page to find the edgeful MCP server, this is the block you need.

• server name: edgeful
• server URL: https://api.edgeful.com/mcp
• transport: streamable HTTP
• auth: OAuth sign-in for Claude, ChatGPT, and GrokBot. clients without OAuth can pass an edgeful API key as a bearer token (config in the setup docs)
• tools: list_allowed_report_endpoints, describe_report_endpoint, call_report_endpoint, get_discovery_scan
• data: computed trading statistics (gap fill, ORB, IB, ADR and ATR, inside bars, previous day's range, and 150+ more) for futures, stocks, forex, and crypto
• report call: a path_template from list_allowed_report_endpoints, path_params with market_type and ticker, and a query with start_date and end_date, plus optional start_time, end_time, timezone, and any report-specific settings from describe_report_endpoint
• discovery scan inputs: market_type, ticker, optional lookback (1mo, 3mo, 6mo, 1y), session, minimum_score
• error without an API-enabled plan: missing_entitlement
• setup docs: https://www.edgeful.com/docs/getting-started/set-up-mcp
• REST endpoint index: https://www.edgeful.com/docs/llms.txt

key takeaways

• the edgeful MCP is one server, https://api.edgeful.com/mcp, for Claude, ChatGPT, and GrokBot. the same reports reach Claude Code, Codex, and Cursor through the edgeful API
• Claude, ChatGPT, and GrokBot take the connector and sign in with your edgeful account. Claude Code, Codex, and Cursor take an API key in a .env file and call the edgeful API directly
• the agent gets 4 tools: list the reports, describe one, run one, and pull the discovery scan
• every plan works. essential sees a starter set, pro sees every report with 12 months of history and per-day rows, all access sees 8 years
• start with 3 prompts: NQ gap fill, ES initial balance, and the ES discovery scan. on the day I ran them, NQ gap downs filled 30 of 53 times (57%) and ES broke only one side of the IB on 89 of 129 days (69%), NY session, last 6 months
• have the agent describe a report before it runs one so the session and defaults match what you meant
• when a question turns into a daily routine, hand it to Claude Code or Cursor and turn it into a dashboard with the edgeful API

edgeful provides historical performance data to help traders make informed decisions. this does not constitute financial advice. past performance is not indicative of future results. all trading involves risk — always do your own analysis and manage your risk accordingly.