What is the ApexVol MCP server?

An MCP (Model Context Protocol) server that gives Claude direct access to ApexVol's options analytics — 43 tools covering IV rank, volatility surfaces, dealer gamma exposure, expected moves, screeners, strategy building and portfolio risk. You ask in plain English; Claude calls the tools and interprets the data.

Can I use the API without Claude?

Yes. The same 59 endpoints are a documented REST API with Bearer-token auth, an OpenAPI 3.1 spec, and a single-file Markdown reference. Anything the MCP tools can do, curl can do.

API and MCP access is in private beta for Pro subscribers. Email support@apexvol.com to join; once enabled you create and manage tokens yourself under Account → API Access.

What data does it use?

Institutional options data covering 5,500+ tickers with 340+ analytics fields per ticker — the same feed behind the ApexVol Terminal. Responses declare their units and data freshness, and quota headers on every response show exactly where you stand.

API & MCP · private beta

Your options desk, inside Claude.

The ApexVol MCP server turns 43 institutional analytics tools into things you can just ask for — IV rank, dealer GEX, expected moves, earnings verdicts, market screens, portfolio stress tests. Prefer code? The same 59 REST endpoints sit behind one Bearer token.

Request beta access

5,500+ tickers · 340+ fields per ticker · Pro plan · tokens self-served, shown once, one-click rotation

Exhibit 01 A real exchange — you ask, Claude calls the tools 43 tools · natural language in, analytics out

You
Is NVDA's earnings straddle overpriced?
Claude · called get_earnings_move_analysis, get_iv_rank
The options market is pricing a ±8.4% move; NVDA's average
                realized earnings move over the last 8 quarters is ±6.1%, and the straddle exceeded
                its implied move only twice. IV rank is 87 — options are rich vs their own history.
                The straddle looks overpriced; premium-selling structures have had the edge here.

Sec. 02 · What you can ask

Plain English in. Desk-grade analytics out.

Every prompt below maps to real tools with real data behind it — no hallucinated numbers.

Volatility

“Is SPY vol rich or cheap right now?”

IV rank & percentile, vol cones, the volatility risk premium through time, term structure, and the ORATS implied-vs-forecast surface.

Positioning

“Where's the gamma wall on QQQ?”

Dealer GEX by strike, charm and vanna exposure, max pain, volume/OI profiles, 0DTE gamma and flip levels.

Earnings

“Should I buy the AMZN earnings straddle?”

Expected vs realized moves, IV crush patterns, post-earnings drift, seasonality, and a combined mispricing verdict.

Screening

“Find me high-VRP names, no earnings this week.”

Eleven preset screens over the full universe — IV rank extremes, steep contango, put skew, vol pairs, pin risk — with filters.

Strategy

“Build an iron condor on SPY and stress it.”

Multi-leg builder with P&L and Greeks, strike optimization, probability of profit, Black-Scholes what-ifs, portfolio scenarios and hedge suggestions.

Deep data

“Pull the chain as it looked before the last selloff.”

Historical EOD chain snapshots, per-strike Greek history, and the raw 340-field ORATS cores row with field selection.

Sec. 03 · Built for agents and engineers

An API that behaves like infrastructure.

The details that matter when a model — or a production system — is on the other end.

Honest data, declared units

IV payloads declare iv_units so nothing gets misread ×100. When upstream data genuinely isn't available, you get an explanation — never silent zeros.

Quotas you can see

Rate-limit and monthly-budget headers on every response, Retry-After on 429s, and a per-endpoint usage breakdown on your account page.

Right-sized for context

Chains default to a near-the-money window that fits an AI context cleanly — widen to the full chain with one parameter when the analysis needs it.

Machine-readable docs

OpenAPI 3.1 spec, a single-file Markdown reference built for pasting next to a CLAUDE.md, and a docs site — all generated from one parity-tested source.

Debuggable failures

Every server error carries a request_id that matches our logs exactly — quote it to support and we're looking at your traceback.

Tokens done right

Self-served, hashed at rest, shown once, 1-year default expiry, one-click rotation — and a --check command that verifies your setup before you touch a config file.

Sec. 04 · Two-minute setup

Connected in three commands.

Exhibit 02 Install → verify → connect Claude Code shown · Claude Desktop uses a JSON config

# 1. Install (pipx keeps it isolated and on your PATH)
pipx install git+https://github.com/ryansilk/apexvol-mcp.git

# 2. Verify your token — prints auth result + remaining quota
APEXVOL_API_TOKEN=avmcp_… apexvol-mcp --check

# 3. Connect Claude Code
claude mcp add apexvol -e APEXVOL_API_TOKEN=avmcp_… -- apexvol-mcp

Rather script it directly? Same token, same endpoints:

curl -H "Authorization: Bearer avmcp_…" \
     "https://apexvol.com/api/mcp/data/iv-rank/NVDA"

Sec. 05 · Access

Private beta, open by request.

API & MCP access runs on the Pro plan. Email us to join the beta — once enabled, you issue and manage your own tokens from your account page, no waiting on support.

Request beta access

Already on the beta? Create a token under Account → API Access.