What is agentic trading?

Using an AI agent (Claude, ChatGPT, or one you build) as your own market analyst instead of following someone else's calls. The agent pulls real data, reasons over it, and lays out the picture; you keep the judgment and place (or skip) the trade. The catch: agents are only as good as the data they can reach. A chatbot with no market data will happily improvise. GammaRips is the data layer that fixes that for options flow. Your agent connects over MCP and reasons over curated, timestamped data with no hindsight in it, instead of vibes.

What does GammaRips actually sell?

Data and tools, not picks. The paid product is MCP access for your AI agent: today's curated options-flow pool in structured form, the opportunity surface for historical setups (how far each contract actually ran, and how far it drew down), a queryable outcome database, regime context, and methodology playbooks: 9 tools in total. Everything human-readable on gammarips.com is free, forever. The $29/mo is the machine connection.

Which AI agents work with this?

Anything that speaks MCP (Model Context Protocol). The free tier works in every MCP client: Claude Code, Codex, Cursor, Gemini CLI, claude.ai and Claude Desktop custom connectors, ChatGPT with Developer mode, Grok connectors, or any MCP client library. The paid tools take a credential two ways. Claude Code, Codex, Cursor and Gemini CLI send your API key in an Authorization header. A chat client that cannot send a header adds https://mcp.gammarips.com/pro instead and signs in with OAuth, so ChatGPT, claude.ai and Grok reach the same tools without a key. Pick your client in the connect section on the homepage. Each tab shows the exact steps and states whether the paid tools run in that client today. Once connected, the 9 tools show up like native capabilities.

Do I need the harness?

No. Any MCP client can call the tools directly. The harness is a free, open-source repo that adds the discipline: it grades tradeability before it considers a story, writes one row for every pool name before any outcome is known, and scores all of them after the close, not only the ones you traded. Clone it from github.com/DevDizzle/gammarips-harness, or ignore it and drive the tools your own way. Its screen uses the paid tools, so the daily loop needs a key.

Why don't you just tell me what to buy?

Two reasons, and we mean both. First, shared picks get crowded: everyone piling into the same thin options contract destroys the very thing they paid for. Second, our own published data shows that mechanically buying everything the engine surfaces, under a fixed exit rule, loses money. The value isn't a pick. It's the surface a good trade comes from: a small curated pool plus the historical data showing how setups like each one actually behaved. We give your agent that surface. It reasons to its own conclusion, sized to your risk, on your horizon.

How is the pool curated?

Every trading night the engine ranks about 3,500 optionable US names by liquidity: the name must have traded 3M+ shares that session and carry a chain with 25 or more listed strikes, then the top 100 by combined chain dollar volume and share volume survive. We keep the bullish names and price one out-of-the-money call in each, chosen on contract liquidity. That is a pool of roughly 40 to 50 contracts, each carrying its flow data, technicals, news context and contract detail. Two safety rails run before our paper cohort's entry: an earnings exclusion and a market-stress check (VIX at or below VIX3M). They screen the cohort's entry, not pool membership, so the pool itself can carry earnings-window names; check any candidate you take. Every field is point-in-time, so nothing your agent sees contains information that wasn't knowable at scan time. Note what this is not: we do not claim these are the best contracts, and we tested that claim. See the next question.

Where's the track record?

On the Track Record page, and in the Lab. Every candidate in the pool is tracked to its real outcome (how high it ran, how far it fell, and a fixed-exit baseline), published as distributions with sample sizes attached, winners and losers counted the same way. We tell you the unflattering part up front: the whole pool bought blindly under a fixed exit is negative. That number is exactly why we sell the data layer and not a pick. The exit is your agent's job.

Is this financial advice?

No. GammaRips is a data vendor. We publish market data, methodology, and research on a paper-trading basis. We never see your account, never manage money, and never make personalized recommendations, and the MCP deliberately has no 'what should I buy' endpoint. What your agent concludes from the data is your analysis, not our advice. If you want personalized investment advice, work with a licensed advisor.

Can't I just scrape the free site?

The webapp shows today's pool and reports in human-readable form, and it always will, free. The MCP is a different animal: structured point-in-time data built for machine reasoning, the historical opportunity-surface and outcome databases that never render on a webpage, exit-rule simulation, regime context, and the methodology playbooks, including the bracket-tournament selection pattern your agent can run against its own objective. You could rebuild some of that from scraping. By the time you have, you'll have built a worse version of the thing we sell for $29.

Evan Parra (founder, ML engineer, data architect) built the engine. The nightly pipeline (scanning, scoring, enriching, publishing) runs autonomously, and every decision is logged to BigQuery. The Lab publishes experiments run on the engine's own data, including the failed ones. Read more on the About page.

What happened to the WhatsApp pick subscription?

Retired. We used to sell a daily pushed pick, and we ended it deliberately. A single shared pick concentrates everyone into one contract, and our own published data keeps showing that buying one shared name under one fixed exit loses money. What varies is how each trader exits, and the exit belongs to the trader. Agent Access buys something better-aligned: full MCP data access, so your agent works the whole pool your way. If you had an active subscription, email evan@gammarips.com and we'll make it right.

What happens if I cancel?

Your agent's MCP access ends with your billing cycle. No retention tricks. Everything human-readable stays free forever: the daily pool, reports, scorecard, methodology, blog, and Lab. Come back whenever your agent misses the data.

Options-flow data for AI agents

Stop asking AI for stock picks.Start giving it real data.

Q: Is the pool better than picking at random?

On returns, we could not show that it is, and we say so. In August 2026 we ran two pre-registered tests, decision rules written before any data was pulled. Against liquidity-matched and random optionable controls over 87 trading days, the pool was not distinguishable on same-day return or on realized excursion. A second test over 57 days found no signal from unusual-activity ranking inside the liquid universe either. Those tests can detect a difference of about 5 percentage points per trade and cannot detect 1 to 2, so the honest reading is no large edge, not no edge. What did measurably improve is whether a contract can be traded at all: on a 60-day window ending 2026-08-14, the share of candidates with no fill at 10:00 ET fell from 40.5% to 6.1% after we moved the scan to liquidity-first. Those are study numbers on a stated window, not a live property of today's pool. This is why we sell the data layer and the outcome history rather than a ranked list.

Every night we rank the US options market by liquidity. Your agent gets a small pool of contracts it can actually trade, plus the history of what setups like them did. Your agent analyzes. You decide.

about 3,500 optionable US stocks ranked nightly · a bullish pool of roughly 40 to 50 calls · no hindsight data · outcomes published, wins and losses

Connect your agent →Browse the pool free

The whole website is free. The free MCP tier needs no card and no key. Agent Access is $29/mo with a 30-day trial.

From nothing to trading, in four steps.

Step one costs nothing and takes about ten seconds. Do that first, and decide whether the data is worth anything to you.

Try the data free

no key, no card

Point any MCP client at the free endpoint. In Claude Code that is one command.

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

Then ask your agent for this morning's brief.

Free tools: pool preview, daily report, regime context, market calendar, and the methodology playbooks.

your agent + gammarips

you ▸ give me this morning's brief.

get_market_calendar_status()get_regime_context()get_daily_report()get_pool()

agent ▸ The market opens on a normal schedule. The VIX term rail passes. Last night's pool holds 44 bullish names, one out-of-the-money call each. Here is the overnight context and the pool preview. Do you want me to work through the names by sector, or by how far out of the money the calls sit?

Illustrative agent session. Not live data, not a recommendation.

Using Cursor, Codex, Gemini CLI, ChatGPT, Claude or Grok instead? Exact steps for every client.

Clone the harness

open source, free

The harness is the daily loop: a screen, a journal, and an after-the-close review. Clone it, then open it in Claude Code or any MCP client that reads .mcp.json.

git clone https://github.com/DevDizzle/gammarips-harness.git
cd gammarips-harness

It is a workflow, not a signal service. The harness is free. The data is what costs money.

See it on GitHub

Subscribe for the pro tools

$29/mo, 30-day trial

Start the trial on the pricing page. Create your key on the account page. It is shown once, so copy it then.

export GAMMARIPS_MCP_KEY="<your key>"

The harness reads that variable from your shell. Never hardcode the key.

Pro tools: liquidity grading, per-name signal detail, outcomes history, contract replay, and the full pool. The harness screens tradeability before it considers a story, so the loop needs them.

A chat client that cannot send a header adds https://mcp.gammarips.com/pro instead and signs in.

Start your 30-day trial

Trade with your agent

Three commands, run in the harness.

/trade    morning: screen the pool, rank a short list or call no-trade
/review   after the close: score every pool name, not only yours
/coach    on demand: your own record, read back to you

Your agent reaches its own contract and its own exit, sized to your risk. There is no pick endpoint, on purpose. Two agents on the same pool should reach different answers.

Paper-traded by default. Educational only, not investment advice.

Wednesday, September 9, 2026

The Partnership Expansion

50contracts in the pool

50bullish names, one call each

A sample of the pool

See every name, free →

Every name in the pool is free to read on this site. The pool is not ranked, and nothing here is a recommendation.

BULL

HPQ

HPQ BULL $32C Sep 18 '26. The institutional flow shows significant call premium concentration at the near-the-money $32 strike, positioning for a swift recovery following the recent 4.6% dip. This opportunistic buying exploits the post-launch dip of HP's new AI-enabled OmniBook and OmniDesk hardware suites, which are expected to drive the ongoing PC refresh cycle. Risk factor: Persistent memory and storage component cost inflation could continue to compress operating margins within the high-volume Personal Systems division, capping near-term equity upside.

BULL

TSLA

TSLA BULL $380C Oct 16 '26. The institutional flow displays a massive concentration of long-term bullish premium targeting the $380 strike over a year out, following analytical upgrades from Goldman Sachs regarding Cybercab's cost-efficiency and Musk's comments on rapid production capabilities. Technically, this bullish flow targets a key zone just below major descending channel resistance, signaling anticipation of a sustained multi-month recovery. Risk factor: Heightened regulatory scrutiny of autonomous taxi fleets and ongoing operating margin pressures from elevated R&D spending could cap structural upside.

BULL

SMR

SMR BULL $12C Sep 18 '26. Institutional flow focused heavily on the near-term $12 strike calls as a rotation back into nuclear and small modular reactor developers lifted the entire complex, including peers like Oklo and the Global X Uranium ETF. Technically, the underlying spot of $11.18 is reclaiming the upper boundary of its multi-week trading range after finding aggressive buyers near the $10 level. Risk factor: The stock remains highly speculative with deep fundamental weakness, including a 99% year-over-year revenue drop in its latest quarter, meaning any sudden cooling in sector momentum could quickly reverse these gains.

BULL

SLB

SLB BULL $60C Oct 16 '26. The institutional flow shows significant accumulation of long-term upside exposure, targeting a return toward 52-week highs following the firm's strategic pivot into data center cooling infrastructure via the Kelvion acquisition and its software integration with S&P Global. This substantial option block positions ahead of multi-quarter accretion from these non-traditional revenue drivers. Risk factor: Sudden downward shifts in international oil exploration budgets could pressure the core oilfield services segment and offset digital-driven valuation gains.

BULL

RIOT

RIOT BULL $23C Sep 18 '26. The institutional flow shows significant premium concentration in the out-of-the-money September 2026 call option, demonstrating strong speculative demand as the underlying trades at $22.26. This aggressive position follows the company's transformative $9.1 billion AI colocation lease and a supportive technical setup near its long-term moving averages. Risk factor: Any sudden correction in Bitcoin pricing or execution delays in the Rockdale AI data center buildout could severely deflate this premium before the contract's expiration.

You can restate the whole rule in one breath.

This runs every trading night, at 23:00 ET.

about 3,500 optionable US stockstraded 3M+ shares that sessionchain carries 25+ listed strikestop 100 by combined liquidity rankbullish names onlyone out-of-the-money call eacha pool of roughly 40 to 50

Two safety rails guard the paper cohort's entry, not this list. The cohort drops any name with earnings in its hold window and stands down when the VIX sits above VIX3M. The pool you browse is not earnings-screened, so check each candidate yourself.

Liquidity decides membership. Not unusual activity. Flow gives your agent context, and it does not choose the pool. We rank the most liquid names and then choose one contract inside each. If we ranked the most liquid contracts instead, you would get SPY and QQQ every day.

We moved the scan to liquidity first because the old funnel picked contracts that were harder to trade. On the 60 trading days ending 2026-08-14, the share of candidates with no fill at 10:00 ET measured 40.5% under the old funnel and 6.1% under this one. Those are study numbers on that window. They are not a property of today's pool, and they are not a claim about returns.

Read the full methodology See how the engine works

Read this before you pay us

If you bought every contract in the pool every morning and closed on one fixed same-day rule (in at the 10:00 ET mark, out at +40% or −30%, flat by 15:45 ET), you would lose money. We know because we tested it, and we publish the record. That is exactly why there is no pick endpoint. The winners sit inside the pool, and the outcome data shows how far each contract ran and how far it fell. Which ones, and how they get traded, is analysis. That is your agent's job. Anyone who sells you a shortcut past that step is selling you a story.

We also tested whether the pool beats random. Over 87 trading days, against liquidity-matched and random optionable controls, we could not show that it does. The test can see a difference of about 5 percentage points per trade, so read that as no large edge rather than no edge. We sell the data layer and the outcome history. We do not sell a ranked list.

I trade my own tool every morning. That is my own process, not a signal to follow. The public receipts are paper-traded and educational only. See the disclosures.

Connect your agent →Read the experiments

Exact steps, for every client

Pick your client. The free tier needs no card, no key, and no signup. Agent Access ($29/mo, 30-day trial) opens the pro tools. Paste a key if your client sends headers. If it cannot, add the pro endpoint and sign in instead. Every tab reaches the same 9 tools.

Free

no card, no key

One command. No card, no key, no signup.

Add the server, then ask Claude for a morning brief.

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

Pro, $29/mo

All paid tools, with your key

All paid tools, and the harness loop. Add the same server with your key.

Mint your key on the account page after the trial starts. Paste it with no spaces and no newline.

claude mcp add --transport http gammarips https://mcp.gammarips.com/mcp \
  --header "Authorization: Bearer YOUR_API_KEY"

2
Or skip the key and sign in. Add the /pro endpoint, then run /mcp inside Claude Code and authenticate.
```
claude mcp add --transport http gammarips https://mcp.gammarips.com/pro
```
3
Or clone the harness: its .mcp.json reads GAMMARIPS_MCP_KEY from your shell.

Start your 30-day free trial, get your key

See all 9 MCP tools →

The Python SDK connects the same way. Steps are on the developer docs.

From the Blog

How the engine builds the pool: methodology, research, and plain-English explainers.

View all →

Stop asking AI for stock picks.Start giving it real data.

From nothing to trading, in four steps.

Try the data free

Clone the harness

Subscribe for the pro tools

Trade with your agent

The Partnership Expansion

A sample of the pool

You can restate the whole rule in one breath.

Read this before you pay us

Exact steps, for every client

Free

Pro, $29/mo

From the Blog

How to Build a Trading Bot with Claude

How to Choose an Options Flow Scanner: What Free Tools Show (and What They Miss)

How to Read Options Flow: A Beginner's Walkthrough

Frequently Asked Questions

One email a week. Catch up in five minutes.

Stop asking AI for stock picks.Start giving it real data.

From nothing to trading, in four steps.

Try the data free

Clone the harness

Subscribe for the pro tools

Trade with your agent

The Partnership Expansion

A sample of the pool

You can restate the whole rule in one breath.

Read this before you pay us

Exact steps, for every client

Free

Pro, $29/mo

From the Blog

How to Build a Trading Bot with Claude

How to Choose an Options Flow Scanner: What Free Tools Show (and What They Miss)

How to Read Options Flow: A Beginner's Walkthrough

Frequently Asked Questions

What is agentic trading?

What does GammaRips actually sell?

Which AI agents work with this?

Do I need the harness?

Why don't you just tell me what to buy?

How is the pool curated?

Is the pool better than picking at random?

Where's the track record?

Is this financial advice?

Can't I just scrape the free site?

Who runs this?

What happened to the WhatsApp pick subscription?

What happens if I cancel?

One email a week. Catch up in five minutes.