Newshot

Open-source harness and MCP dataset for options trading with AI agents

Original text(via HN AI)

gammarips-harness

An open-source agent workflow harness for the GammaRips MCP curated options dataset.

GammaRips is a data layer, not a signal service. Every trading day, the engine filters ~3,500 optionable US stocks down to a curated cohort of liquid, tradable, high-edge option contracts and serves that dataset over MCP. Your agent pulls the surviving cohort, layers its own live catalyst research (X/Twitter cashtags, filings, web search, parallel subagents), and builds a deterministic trade plan with historical MFE/MAE excursion data.

The Product: Curated Options Dataset via MCP

Instead of dumping noisy raw chains into an LLM context window, GammaRips filters the US options universe down to the daily cohort of surviving contracts that pass strict quantitative and liquidity gates:

  • Mid Price: Mid >= $1.00
  • Open Interest: OI >= 500
  • Delta Band: Delta 0.22–0.48
  • Days to Expiration: DTE 10–45
  • Volatility Edge: RV/IV >= 0.85 (realized volatility vs. implied volatility)

MCP Endpoints

Connect any MCP-compatible agent (Claude Code, Grok, Codex, Cursor, or custom runners):

  • Pro (OAuth): https://mcp.gammarips.com/pro
  • API Key (Bearer): https://mcp.gammarips.com/mcp with header Authorization: Bearer gr_live_...
  • Free Preview (No Key): https://mcp.gammarips.com/mcp (serves the daily report, regime context, market calendar, playbooks, and cohort preview)

3-Step Agent Workflow

1. Pull the Daily Cohort (get_pool)

Fetch today's surviving tradable option contracts and their point-in-time volatility and liquidity metrics (rv_iv_ratio, open_interest, mid_price, delta, dte, and opportunity surfaces), alongside market calendar status (get_market_calendar_status) and VIX term-structure regime (get_regime_context).

2. Layer Your Own Catalyst Search (Bring Your Own X / Web Agent)

GammaRips handles the quantitative options filtering; you bring the catalyst layer. Have your agent (Claude Code, Grok, Codex, Cursor) spin up parallel subagents or live X/Twitter and web search — for example, Grok searching live $TICKER cashtag chatter, SEC/FDA regulatory filings, or unpriced catalysts — to debate and rank the surviving contracts.

3. Build the Trade Plan / Execute

Size the position against your risk budget, set deterministic bracket exits (profit target, stop loss, and mandatory time-stop) using historical MFE/MAE excursion distributions (query_outcomes and replay_contract), pre-register the decision in eval/funnel-log.jsonl, and optionally route the order to your own broker MCP.

Quick Start

Try the free preview (no key, 10 seconds)


        claude mcp add --transport http gammarips https://mcp.gammarips.com/mcp
      

Ask your agent for today's brief to inspect the market calendar, VIX regime context, daily report, and cohort preview.

Real recording: connecting to the GammaRips MCP and reading the daily report and regime context.
Real recording: connecting to the GammaRips MCP and reading the daily report and regime context.

Connect with Pro (OAuth or API Key)

  • Subscribe at gammarips.com/pricing and create an API key at gammarips.com/account (or authenticate directly via OAuth at https://mcp.gammarips.com/pro).
  • Export your key for .mcp.json: export GAMMARIPS_MCP_KEY="gr_live_..."

        export GAMMARIPS_MCP_KEY="gr_live_..."
      
  • Or install as a Claude Code plugin: /plugin marketplace add DevDizzle/gammarips-harness /plugin install gammarips@gammarips

        /plugin marketplace add DevDizzle/gammarips-harness
/plugin install gammarips@gammarips
      
  • Run the daily skills: /trade Step 1–3 morning workflow: pull cohort, layer catalysts, build exit plan /review After the close: score every cohort contract with MFE/MAE outcomes /coach On demand: referee trading behavior against your written rules

        /trade    Step 1–3 morning workflow: pull cohort, layer catalysts, build exit plan
/review   After the close: score every cohort contract with MFE/MAE outcomes
/coach    On demand: referee trading behavior against your written rules
      

Repository Layout

PathDescription
CLAUDE.mdAgent instructions: dataset specification, 3-step workflow, and core rules
.mcp.jsonStreamable HTTP MCP config ( https://mcp.gammarips.com/mcp with Bearer ${GAMMARIPS_MCP_KEY} )
docs/TRADING-DOCTRINE.mdRisk rules, hard exclusions, sizing framework, and deterministic exit discipline
docs/OPERATIONS.mdMCP endpoints ( /pro OAuth and /mcp Bearer key), session timing, and failure handling
.claude/skills/Agent procedures: trade , review , coach , wiki-distill , ste100
wiki/Claim-tagged research notes ( findings/ , literature/ ) and index registries
eval/Your local dataset: funnel-log.jsonl , trades.jsonl , behavior-ledger.jsonl
scripts/Validators ( funnel_log.py , behavior_log.py , lint.py ) and evaluators ( edge.py , skill_eval.py )

Not Investment Advice

GammaRips is a financial data vendor. Everything in this repository is for educational and research purposes, configured for paper trading by default. Nothing here constitutes investment advice. Trade your own capital at your own risk.

The text above was extracted automatically from the source page and may miss images, tables or code blocks. The original is authoritative.

Published by HN AI
This site only summarises and aggregates; copyright stays with the original author. Headlines and summaries follow the source — where they differ, the original wins.
Read the original ↗