Kalshi-nativeCLIfree
kalshi-trading-bot-cli: what it does, what it costs to run, where it stops
Last checked 2026-09-08
kalshi-trading-bot-cli is a Bun-based command line tool from OctagonAI that researches Kalshi markets, produces an independent probability estimate, computes edge as the gap between model price and the live order book, sizes with half-Kelly and places the order.
It is Kalshi-only and Octagon-centric. The research, edge scanning, semantic search, clustering and basket commands all route through the Octagon Research API when a key is set.
What it does
The command set covers search, analyze, watch, buy, sell, cancel, portfolio and backtest, plus a discovery layer: similar, clusters, peers, correlate and basket build. Every command takes a json flag for scripts and agents.
Backtest scores the model against Kalshi settlement for resolved markets and mark-to-market for open ones, reporting Brier scores, hit rate and flat-bet ROI.
- bunx kalshi-trading-bot-cli@latest starts a setup wizard on first run
- Five risk gates: Kelly, liquidity, correlation, concentration, drawdown
- Defaults: kelly_multiplier 0.5, max_drawdown 0.20, max_positions 10, max_per_category 3, daily_loss_limit $200
- KALSHI_USE_DEMO=true routes everything to Kalshi's demo environment
Strategy
Edge equals model probability minus market price. The analyze command returns the model probability, ranked price drivers and a confidence label, then recommends a contract count at the current ask. The example in the README shows a 72% model against a 58% market, a 14-point edge, and a 3-contract YES recommendation.
The basket commands push universe selection, cluster caps, pairwise correlation gates and Kelly sizing server-side into one call, which is aimed at agents rather than a person typing tickers.
Setup and requirements
Bun 1.1 or newer is required; Node.js will not work because the tool uses bun:sqlite. You need a Kalshi API key ID and RSA private key PEM, one LLM provider key, and an Octagon key for the research features. The default model is GPT-5.4, so OPENAI_API_KEY is required unless you switch providers.
Config, cache and the SQLite database live in ~/.kalshi-bot/. Anonymous telemetry is on by default and turned off with TELEMETRY_ENABLED=false.
Activity and maintenance
383 stars and 105 forks as of 2026-09-08. Created 2025-07-14, last push 2026-08-22, 12 open issues. MIT licensed, TypeScript. The homepage is octagonai.co/markets.
Where it falls short
The useful parts need paid Octagon credits at 3 per report, with a default daily ceiling of 100. Without an Octagon key, search falls back to a local cache and analyze loses its research.
Kalshi only, US users only in practice. The README documents bunx quirks (install chatter in JSON output, parallel-invocation races) that you must work around for scripting.
Who should use it
A Kalshi trader who wants a research-to-order pipeline in a terminal and is fine paying Octagon for the probability model. Start with KALSHI_USE_DEMO=true.
Not for cross-venue work and not for anyone who wants the model to be their own.
Questions about kalshi-trading-bot-cli
Is kalshi-trading-bot-cli free?
The CLI is MIT licensed. Octagon research costs 3 credits per report and the LLM provider bills separately. Kalshi itself charges its own trading fees.
Does kalshi-trading-bot-cli work with Polymarket?
No. It is Kalshi-only.
Does kalshi-trading-bot-cli place real trades?
Yes, buy and sell place orders through your Kalshi API key. Set KALSHI_USE_DEMO=true to simulate.
Do I need Node.js for kalshi-trading-bot-cli?
No, you need Bun 1.1 or newer. The README says Node.js will not run it.
Bots die in execution. Terminals are execution.
Wanted a bot for speed and cross-venue vision? The free Kairos terminal trades Kalshi, Polymarket and Predict.fun from one book, no keys to manage.