Add Alphanume's Financial Data MCP to Claude

Connect Claude or Codex to Alphanume's financial datasets, verify the connection, and run a three-dataset market screen without copying API responses into chat.

You can give Claude direct access to Alphanume's financial data in about one minute. Add https://mcp.alphanume.com/mcp as a custom connector, sign in with your Alphanume account, and the same connection becomes available across Claude web, desktop, mobile, Cowork, and Claude Code. There is no API key to copy into Claude and no header to configure.

The connector exposes 25 read-only dataset and status tools, including Next-Day Movers, IV/HV Premium, and the S&P 500 Risk Regime. This is useful when a research question spans datasets. Instead of downloading three files, aligning dates by hand, and pasting partial tables into a model, you can ask for the workflow in plain English and require the model to show the dates, fields, and missing values it used.

What the Alphanume MCP connection provides

MCP is the interface between the model client and Alphanume's hosted tools. The server is a thin proxy over the REST API rather than a separate data product. Each tool maps to a dataset and returns an envelope shaped as { "count": N, "data": [...] }. Tool descriptions tell the model which filters are available and what each field means.

For this example, get_next_day_movers returns a daily shortlist of equities a volatility model expects to make an outsized move in the following session. get_iv_hv_premium compares roughly 30-day implied and realized volatility and includes their spread, ratio, and cross-sectional ranks. get_sp500_risk_regime returns a daily risk-on or risk-off label designed as an overlay for other research. The Next-Day Movers docs, IV/HV Premium docs, and Risk Regime docs describe the underlying datasets before you ask a model to combine them.

Connect Claude in one minute

Use the OAuth connector for the shortest setup. It matches the tools to the Alphanume account you already use for your subscription, so your existing tier carries through automatically:

1. Open Claude on the web or desktop, go to Settings, then Connectors.
2. Select Add custom connector, name it Alphanume, and paste https://mcp.alphanume.com/mcp.
3. Complete the browser sign-in with the same Alphanume account you subscribe with.
4. Open Claude Code while signed in to the same Claude account. Run /mcp and confirm that alphanume appears as connected.
5. Ask Claude to call check_api_status and report the returned status without making any market claim.

The connector added through Claude propagates to Claude Code, so you do not need to register it a second time at the command line. If the server appears but the status call fails, confirm that the browser sign-in used the intended Alphanume account before changing the prompt. The Alphanume MCP documentation and official Claude Code MCP documentation cover the connector and client controls.

The API-key route is a fallback for clients or environments that cannot complete a browser sign-in. Obtain a key from the pricing and access page, keep it out of source control and prompts, and point the client at https://mcp.alphanume.com/key/mcp. For Claude Code, the fallback command is:

claude mcp add alphanume-key --transport http https://mcp.alphanume.com/key/mcp --header "X-API-Key: alp_YOUR_KEY"

Use the same server in Codex

Codex also supports the OAuth endpoint as a streamable HTTP server. Add this entry to ~/.codex/config.toml, or to a trusted project's .codex/config.toml if the connection should be project-scoped:

[mcp_servers.alphanume]
url = "https://mcp.alphanume.com/mcp"

Restart the Codex host, run codex mcp login alphanume, and complete the Alphanume browser sign-in. Then run codex mcp list in the CLI and use /mcp in the terminal interface to confirm the server is active. The official Codex MCP documentation is the source of truth for the current configuration fields and client controls.

If OAuth cannot run in the environment, use the explicit API-key fallback at https://mcp.alphanume.com/key/mcp. Codex supports an environment-backed header, which keeps the value out of the TOML file:

[mcp_servers.alphanume_key]
url = "https://mcp.alphanume.com/key/mcp"
env_http_headers = { "X-API-Key" = "ALPHANUME_API_KEY" }

Run a three-dataset research prompt

Once check_api_status succeeds, use a prompt that names the research sequence and the evidence required. This makes the output easier to audit than a broad request such as "find me a trade":

Use only Alphanume MCP tools for the data in this task.

1. Call check_api_status.
2. Get the latest available S&P 500 Risk Regime row.
3. Get the latest available Next-Day Movers list.
4. For every mover, get the latest settled IV/HV Premium row on or before the mover list date by setting only_final=true.
5. Return one table with mover date, ticker, risk regime date and label, IV, HV, IV/HV ratio, ratio rank, is_final, realized return, and absolute move.
6. Keep null values as null. Separate missing matches from the main table. Do not infer direction, expected return, or a trade recommendation.
7. After the table, list the exact Alphanume tools and filters you used.

The prompt creates a reproducible join rather than asking the model to improvise a narrative. The movers list narrows the universe, IV/HV Premium supplies options-pricing context for each name, and Risk Regime adds a market-level label. Requiring source dates prevents the model from silently treating records from different sessions as simultaneous. Requiring settled volatility rows avoids mixing intraday estimates with end-of-day observations.

Read the result without overstating it

These three datasets answer different questions. Next-Day Movers estimates movement magnitude, not direction. A ticker appearing in the list is not a bullish or bearish forecast. A high IV/HV ratio says options are rich relative to recent realized volatility, but it does not prove that selling options has positive expected value after earnings risk, skew, spreads, and transaction costs. The Risk Regime is a binary overlay, not a timing system or a substitute for position-level risk controls.

Freshness also differs by dataset. The newest Next-Day Movers rows can have null return and absolute_move values until the following session trades and the outcome is filled. IV/HV Premium updates intraday with is_final=0 and settles after the close with is_final=1. A model should preserve those states, not fill them from nearby dates or describe a provisional row as final.

Know the access and tool limits

Free access currently covers the trailing 20 trading sessions and is delayed by one trading session. Pro access provides current data and full history. A range-access error means the requested period is outside the account's tier, not that the dataset has no records. Dataset tools default to a 500-row response cap, so historical work should use narrow date and ticker filters and inspect any truncation marker.

• MCP makes tool access conversational, but it does not validate a research hypothesis for you.
• The model can choose a wrong filter or join incompatible dates, so require an audit trail in the prompt.
• The hosted MCP server forwards the caller's API access and does not expand the underlying subscription tier.
• Do not paste a live API key into the conversation or commit it in a project configuration file.

Build the next research loop

After the first screen works, save the prompt beside the resulting table and rerun it on later dates. That gives you a small point-in-time research log: what the movers list contained, how options were priced, which regime label was active, and which outcomes were still unresolved. For a broader framework on delegating data work while keeping the model's claims testable, read Using Claude for Trading Research. Then use the dataset documentation above to tighten filters before turning the workflow into a backtest or scheduled report.