mcpbeat

Kalshi API

agiprolabs/kalshi-api

Kalshi exchange mechanics — RSA-PSS auth, order schema, YES/NO orderbook convention, WebSocket, and endpoint surface. Market-type-agnostic shared layer for all Kalshi skills.

6k tokens
context cost
the whole folder, loaded on every use
5
files
ships runnable scripts
0
copies elsewhere
how many repositories repackaged it
257
stars on the repo
on the repository, not the skill itself

Install

one command, takes just this skill from the repository
npx skills add https://github.com/agiprolabs/claude-trading-skills --skill kalshi-api

The instruction itself

13 sections, as written by the author

Kalshi API

CFTC-regulated US event exchange. USD-denominated binary contracts settle at $1.00 (YES wins) or $0.00 (NO wins). REST + WebSocket, RSA-PSS authentication on every request.

For contract semantics and settlement rules, see the kalshi-weather-markets and kalshi-crypto-index-markets skills. For strategy, sizing, and backtesting, see prediction-market-strategy.


> VERIFY BEFORE CODING.

> The Kalshi API has broken backward compatibility before: the host changed (old trading-api.kalshi.com → dead), and the order schema changed (integer cents → dollar strings). Always smoke-test signing and order bodies against a live response before shipping.

>

> Canonical sources:

> - API reference: <https://docs.kalshi.com> (legacy mirror: <https://trading-api.readme.io>)

> - Official Python starter: <https://github.com/Kalshi/kalshi-starter-code-python>


Overview

  • Base URL: https://api.elections.kalshi.com/trade-api/v2
  • Auth: RSA-PSS on every request — there are no public/unauthenticated endpoints
  • No demo parity: the demo environment (demo-api.kalshi.co) has a near-empty book; use production even for read-only pulls
  • Contracts: $0.01–$0.99 per contract; pay price if YES wins, lose price if NO wins; max payout = $1.00

Quick Start

1. Credentials

KALSHI_KEY_ID=<your-key-uuid>
KALSHI_PRIVATE_KEY_PATH=~/.kalshi/private.pem

Generate the key in the Kalshi dashboard. Store secrets in environment variables or a secrets manager — never in code.

2. Install

pip install httpx cryptography

3. Host + auth (the part everyone gets wrong)

The host and signature format are where implementations break. Three common failures:

  • Using the old trading-api.kalshi.com host → 401
  • Including the query string in the signed path → 401
  • Signing with seconds instead of milliseconds → 401
import os, time, base64, httpx
from cryptography.hazmat.primitives import hashes, serialization
from cryptography.hazmat.primitives.asymmetric import padding

BASE = "https://api.elections.kalshi.com/trade-api/v2"
KEY_ID = os.environ["KALSHI_KEY_ID"]
with open(os.environ["KALSHI_PRIVATE_KEY_PATH"], "rb") as f:
    PRIV = serialization.load_pem_private_key(f.read(), password=None)

def _headers(method: str, path: str) -> dict:
    """path must include /trade-api/v2 prefix and exclude query string."""
    ts = str(int(time.time() * 1000))             # milliseconds
    msg = f"{ts}{method}{path}".encode()
    sig = PRIV.sign(
        msg,
        padding.PSS(mgf=padding.MGF1(hashes.SHA256()),
                    salt_length=padding.PSS.DIGEST_LENGTH),
        hashes.SHA256(),
    )
    return {
        "KALSHI-ACCESS-KEY": KEY_ID,
        "KALSHI-ACCESS-TIMESTAMP": ts,
        "KALSHI-ACCESS-SIGNATURE": base64.b64encode(sig).decode(),
    }

def get(path: str, params=None):
    # Sign the path only — query goes into params, not the signature
    r = httpx.get(BASE + path, params=params,
                  headers=_headers("GET", "/trade-api/v2" + path))
    r.raise_for_status()
    return r.json()

def post(path: str, body: dict):
    r = httpx.post(BASE + path, json=body,
                   headers=_headers("POST", "/trade-api/v2" + path))
    r.raise_for_status()
    return r.json()

# Example: open markets in a series
markets = get("/markets", params={"series_ticker": "KXHIGHNY", "status": "open"})

Signature spec: RSA-PSS, MGF1 over SHA-256, salt length = PSS.DIGEST_LENGTH. String to sign: {timestamp_ms}{METHOD}{path} where path includes /trade-api/v2 and excludes the query string.

Headers: KALSHI-ACCESS-KEY (UUID), KALSHI-ACCESS-TIMESTAMP (ms), KALSHI-ACCESS-SIGNATURE (base64).

4. Candlestick history

# Returns OHLC for yes_bid / yes_ask + volume + open_interest
# Values are dollar strings: {"close": "0.42"}
candles = get(
    f"/series/KXHIGHNY/markets/{ticker}/candlesticks",
    params={"start_ts": start_epoch, "end_ts": end_epoch, "period_interval": 60},
)

period_interval is in minutes: 1, 60, or 1440.


YES/NO Order-Book Convention

On Kalshi, yes and no are both resting BID ladders — there is no separate ask book. To take the other side you lift the opposing bid:

no_ask  = 1 − best_yes_bid    # cost to buy NO right now (lift YES bids)
yes_ask = 1 − best_no_bid     # cost to buy YES right now (lift NO bids)
P(YES) mid = (best_yes_bid + (1 − best_no_bid)) / 2

Getting this backwards silently inverts every signal. Use the helpers in scripts/kalshi_orderbook.py.

Orderbook response comes in two variants depending on API tier — normalize before using:

{"orderbook": {"yes": [[price, size], ...], "no": [[price, size], ...]}}

If a price value is > 1.0, it is integer cents — divide by 100.


Order Schema

POST /trade-api/v2/portfolio/orders uses fixed-point dollar STRINGS, not integers. The old schema (integer cents, count, yes_price) returns 400 invalid_parameters.

{
  "ticker": "KXHIGHNY-26JUN02-B75.5",
  "action": "buy",
  "side": "yes",
  "count_fp": "1.00",
  "yes_price_dollars": "0.01",
  "client_order_id": "my-strategy-001",
  "time_in_force": "good_till_canceled"
}

Critical field rules — each violation returns 400:

| Field | Rule | Common mistake that 400s |

|---|---|---|

| count_fp | fixed-point string "1.00" | integer count: 1 |

| {side}_price_dollars | dollar string "0.01" | integer cents yes_price: 1 |

| time_in_force | required: good_till_canceled \| immediate_or_cancel \| fill_or_kill | omitted |

| client_order_id | [A-Za-z0-9-] only | . or : in the string — bracket tickers contain ., so never copy the ticker directly |

| type | do not send | "type": "limit" |

For the full order lifecycle (amend, decrease, cancel, batch) and strike_type gotchas, see references/auth-and-orders.md.


Endpoint Summary

| Category | Endpoint | Notes |

|---|---|---|

| Balance | GET /portfolio/balance | — |

| Positions | GET /portfolio/positions | — |

| Orders | GET /portfolio/orders | ?status=resting\|canceled\|executed |

| Place order | POST /portfolio/orders | dollar-string schema above |

| Cancel | DELETE /portfolio/orders/{id} | returns {"order": {...status: "canceled"}} |

| Amend | POST /portfolio/orders/{id}/amend | ticker required in body |

| Decrease | POST /portfolio/orders/{id}/decrease | {"reduce_by_fp": "1.00"} |

| Batch | POST /portfolio/orders/batched | {"orders": [...]} |

| Fills | GET /portfolio/fills | ?limit=N |

| Settlements | GET /portfolio/settlements | ?limit=N |

| Markets | GET /markets | ?series_ticker=&status=open&limit=500 |

| Orderbook | GET /markets/{ticker}/orderbook | ?depth=N |

| Candlesticks | GET /series/{series}/markets/{ticker}/candlesticks | ?start_ts=&end_ts=&period_interval=60 |

| Trades | GET /markets/trades | recent trade prints |

Full endpoint surface, market metadata fields, and rate limits: references/endpoints-and-marketdata.md.

WebSocket discovery pipeline: references/websocket.md.


Files

References

  • references/auth-and-orders.md — RSA-PSS spec, dollar-string order schema, time_in_force, client_order_id sanitization, amend/decrease/cancel lifecycle, strike_type gotcha, fees
  • references/endpoints-and-marketdata.md — Full endpoint table, orderbook variants, market metadata fields (result, open_time, close_time, series_ticker), candlesticks, rate limits
  • references/websocket.md — WS host, discovery pipeline, channels, signing the WS upgrade

Scripts

  • scripts/kalshi_orderbook.py — YES/NO bid-ladder helpers (no_ask, yes_ask, p_yes_mid, overround, kalshi_fee). Pure stdlib, no keys, runs offline.

How to use it

Copy the folder

Take agiprolabs/kalshi-api from the repository into ~/.claude/skills for personal use, or into .claude/skills inside a project.

Check the name does not clash

The agent identifies a skill by the name field in its header. Two skills with the same name cannot sit side by side — one of them will be ignored.

Install what it needs

The instructions reference pip. Without those the skill loads but fails at the first command.