Built for prediction markets

Trading bots for Polymarket and Kalshi, built to spec.

We design and ship custom prediction-market trading bots — for traders, prop firms, and curious operators. The code is yours at delivery. Backtested before live, every time. No profit promises.

Polymarket CLOB API Kalshi Trade API Polygon · Base · Ethereum v2026.05
Three doors

Pick the entry that fits you.

SF-001 · entry-points
What we build

Eight bot archetypes. Or your own.

Every project starts from the spec, not a template. These are the patterns we've shipped most often — pick a starting point or describe yours.

SF-002 · services
Anatomy of an edge

When YES + NO < $1.00, the bot fires.

Every binary market resolves to exactly $1.00. So when the two orderbooks momentarily mis-quote — YES at $0.49, NO at $0.49 — there's a guaranteed 2¢ to capture. In practice: 3–7 cents per fill, ~120 opportunities/week on liquid markets in 2026.

SF-005 · live snapshot
Trump approval > 45% · YES book YES
$0.521,840
$0.511,200
$0.50680
$0.49420
— spread $0.01 —
$0.48540
$0.47980
$0.461,600
YES.ask + NO.ask = $0.98 2¢ edge · 2.04% gross → FIRE @ 420 · <300ms
Trump approval > 45% · NO book NO
$0.521,620
$0.51920
$0.50600
$0.49420
— spread $0.02 —
$0.47660
$0.461,150
$0.451,720

Illustrative snapshot · prices and sizes simulated for clarity · real arb gaps are typically 0.3¢–2¢

Why automation, why now

Prediction markets matured. Most traders haven't.

In 2026, the surface area is large enough that manual execution leaves real money on the table. The bottleneck isn't ideas — it's infrastructure.

SF-003 · context
$12.4B
Polymarket monthly trading volume, Jan 2026
src: Datawallet
~100k
Polymarket daily active users
src: Polymarket dashboard
100 req/min
CLOB public endpoint rate limit — priority queue required at scale
src: Polymarket docs
<$0.005
Polygon gas cost per trade — ~$5 / 1,000 trades infra burn
src: Polygonscan, 30d median
Our process

Six steps. Discovery to optimize.

Nothing touches real capital before it survives a 6–12 month backtest and two weeks of paper trading. If the backtest is red, we don't deploy. We tell you, we revise the strategy, or we refund.

SF-004 · workflow
01 · DiscoveryTranslate your ideaIf you can describe what you want, we turn it into a technical spec. Even if you're not a trader.
02 · StrategyDefine edge & riskWe document the strategy assumptions, position sizing, drawdown caps, and exit conditions.
03 · Backtest6–12 months of dataSurvive historical conditions including the bad ones. Red backtest = we don't ship.
04 · BuildCLOB-native codePython or TypeScript. Direct CLOB API. Type-checked. Monitored. Reviewed by a second engineer.
05 · DeployPaper, then liveTwo weeks of paper trading minimum. Capital is wired only after paper matches the backtest.
06 · OptimizeMaintenance retainerMonitoring, API drift fixes, strategy tuning. Optional. The repo is yours either way.
Recent work

Three shipped projects, three different shapes.

All case studies
The contract

What you get vs. what we don't promise.

Honest framing protects both sides. Here's exactly where the line sits.

YOU GET Deliverables, contractual

  • Full source repository. Git access transferred at delivery. Branches, history, secrets template.
  • Technical documentation. Architecture diagram, runbook, every config flag explained.
  • Backtest report. 6–12 months historical, with assumptions, drawdown chart, sensitivity analysis.
  • Infrastructure keys. AWS/Hetzner/Railway — your account, your billing, your control.
  • Monitoring stack. Grafana dashboards + alerting to your channel of choice (Discord, Telegram, PagerDuty).
  • A 30-day warranty. Bugs in shipped code are fixed at no charge for the first 30 days post-deploy.

WE DON'T Promise these things

  • Guaranteed returns. If someone promises 10% monthly, it's a scam. We deliver a system, not a yield.
  • "Passive income." Bots require monitoring. Markets shift. Strategies decay. Plan for upkeep.
  • Profit on day one. Paper trading first. Capital wires only after live matches backtest.
  • ToS workarounds. We won't help bypass geographic restrictions (Polymarket § 2.1.4).
  • A black box. If you don't understand a line of the strategy code, we explain it before you wire funds.
  • Hidden recurring fees. No revenue share. No subscription. Optional retainer is opt-in only.
Real code, not screenshots

What we ship looks like this.

No abstraction layer between you and the CLOB. Strategy logic stays readable — fewer than 200 lines for most archetypes — so any senior engineer you hire later can pick it up in a day.

  • Direct CLOB calls. No SaaS wrapper, no rate-limit surprises hidden in a vendor's quota.
  • Strategy is data. Thresholds, sizing, filters live in YAML — you tune without redeploying.
  • Every event logged. If a trade fires you didn't expect, the reason is pinned to a config line.
bot/arb_btc15m.py python
from py_clob_client.client import ClobClient

# Connect to Polymarket CLOB
client = ClobClient(
    host="https://clob.polymarket.com",
    key=PRIVATE_KEY, chain_id=137,
)

# Watch BTC 15m markets for sum-to-one arbitrage
for mkt in client.get_markets(category="crypto"):
    yes = client.get_orderbook(mkt.token_yes)
    no  = client.get_orderbook(mkt.token_no)

    if yes.best_ask + no.best_ask < 0.98:
        execute_arbitrage(yes, no, size=position_size)
        log.info("arb fired", edge=0.02)
!
Compliance note.

We build bots only for users in jurisdictions where Polymarket and Kalshi are accessible. Polymarket restricts US users (§ 2.1.4 ToS); Kalshi is CFTC-regulated and US-available. We will not assist with VPN, multi-account, or KYC-circumvention setups. If your use case requires that, we are not the team.

Ready when you are

Tell us what you want the bot to do.

The first conversation is free, 30 minutes, no pitch. If the project doesn't make sense, we'll tell you. If it does, you'll have a written spec and a quote within five business days.