kalshi-sdk 2.0.0

A professional Python SDK for the Kalshi prediction markets API

kalshi-sdk

A professional, spec-first Python SDK for the Kalshi prediction markets API.

• Full coverage of the Kalshi REST API (89 endpoints across 19 resources) and WebSocket API (11 channels).
• Sync and async clients sharing one transport — no thread-pool wrapping.
• Typed end-to-end: Pydantic v2 models, mypy --strict clean, ships py.typed. Literal types on fixed-enum kwargs.
• Spec-aligned with drift guards: hard-fail contract tests catch query, body, and WebSocket payload drift on every commit.
• Safe defaults: only idempotent verbs (GET/HEAD/OPTIONS) retry; POST/DELETE never retry to avoid duplicate orders or cancels.
• DataFrame-ready: optional pandas / polars extras for analysis workflows.
• Offline-testable: record/replay mock transport (kalshi.testing) for SDK consumers building integration tests.

📖 Full documentation: https://texascoding.github.io/kalshi-python-sdk/

Install

pip install kalshi-sdk

Requires Python 3.12+.

Quickstart — sync

from kalshi import KalshiClient

with KalshiClient(
    key_id="your-key-id",
    private_key_path="~/.kalshi/private_key.pem",
) as client:
    page = client.markets.list(status="open", limit=10)
    for market in page:
        print(market.ticker, market.yes_bid, market.yes_ask)

Quickstart — async

import asyncio
from kalshi import AsyncKalshiClient

async def main() -> None:
    async with AsyncKalshiClient(
        key_id="your-key-id",
        private_key_path="~/.kalshi/private_key.pem",
    ) as client:
        # list_all() yields across pages — works directly with `async for`.
        async for market in client.markets.list_all(status="open"):
            print(market.ticker, market.yes_bid)

asyncio.run(main())

Authentication

Kalshi uses RSA-PSS request signing. Generate a key pair in your Kalshi account settings and download the PEM.

From environment variables

export KALSHI_KEY_ID="..."
export KALSHI_PRIVATE_KEY_PATH="~/.kalshi/private_key.pem"
# or, inline:
export KALSHI_PRIVATE_KEY="-----BEGIN PRIVATE KEY-----..."

# Optional:
export KALSHI_DEMO=true              # use the demo (sandbox) environment
export KALSHI_API_BASE_URL=...       # override base URL

from kalshi import KalshiClient

client = KalshiClient.from_env()

from_env() returns an unauthenticated client if no credentials are set. Public endpoints still work; private endpoints raise AuthRequiredError.

Demo vs production

KalshiClient(key_id="...", private_key_path="...", demo=True)   # sandbox
KalshiClient(key_id="...", private_key_path="...")              # production (default)

Public / unauthenticated usage

You don't need credentials to read public market data:

from kalshi import KalshiClient

with KalshiClient(demo=True) as client:
    assert client.is_authenticated is False
    markets = client.markets.list(status="open", limit=5)

Placing orders

from kalshi import KalshiClient

with KalshiClient.from_env() as client:
    order = client.orders.create(
        ticker="EXAMPLE-25-T",
        side="yes",
        action="buy",
        count=10,
        yes_price="0.65",          # 65 cents
        time_in_force="good_till_canceled",
        client_order_id="my-uuid", # idempotency key
    )
    print(order.order_id, order.status)

Prices are decimal dollars (e.g. "0.65") per the Kalshi spec. Internally the SDK uses Decimal via the DollarDecimal type — never float.

Every POST/PUT/DELETE-with-body method also accepts a pre-built request model as an alternative to individual kwargs (useful for programmatic order construction):

from kalshi import CreateOrderRequest

client.orders.create(request=CreateOrderRequest(
    ticker="EXAMPLE-25-T", side="yes", action="buy",
    count=10, yes_price="0.65",
))

WebSocket streaming

import asyncio
from kalshi import KalshiAuth, KalshiConfig
from kalshi.ws.client import KalshiWebSocket

async def main() -> None:
    auth = KalshiAuth.from_key_path("your-key-id", "~/.kalshi/private_key.pem")
    config = KalshiConfig.demo()  # or KalshiConfig.production()

    ws = KalshiWebSocket(auth=auth, config=config)
    async with ws.connect() as session:
        stream = await session.subscribe_orderbook_delta(tickers=["EXAMPLE-25-T"])
        async for msg in stream:
            print(msg)

asyncio.run(main())

Available channels (11): ticker, trade, orderbook_delta, fill, market_positions, user_orders, order_group, market_lifecycle, multivariate, multivariate_lifecycle, communications.

Error handling

All SDK errors inherit from KalshiError:

from kalshi import (
    KalshiError,
    KalshiAuthError,        # 401 / 403
    AuthRequiredError,      # called private endpoint without credentials
    KalshiNotFoundError,    # 404
    KalshiValidationError,  # 400 (has .details: dict[str, str])
    KalshiRateLimitError,   # 429 (has .retry_after: float | None)
    KalshiServerError,      # 5xx
    # WebSocket-specific:
    KalshiWebSocketError,
    KalshiConnectionError,
    KalshiSequenceGapError,
    KalshiBackpressureError,
    KalshiSubscriptionError,
)

try:
    client.markets.get("DOES-NOT-EXIST")
except KalshiNotFoundError as e:
    print(e.status_code, str(e))

Retry policy

• Retries on 429, 502, 503, 504, 500 (idempotent GET only).
• POST and DELETE are never retried — duplicate order / cancel risk.
• Exponential backoff with jitter, capped at retry_max_delay.
• Retry-After is honored but capped at retry_max_delay to prevent a server-controlled stall.

Tune via KalshiConfig:

from kalshi import KalshiClient, KalshiConfig

config = KalshiConfig(
    timeout=10.0,
    max_retries=5,
    retry_base_delay=0.5,
    retry_max_delay=15.0,
    # Connection pool / HTTP-2 tuning (opt-in; defaults preserve v1 behavior)
    http2=False,
    limits=None,  # httpx.Limits(max_connections=..., keepalive_expiry=...)
    extra_headers={"X-My-Tag": "foo"},
)
client = KalshiClient(key_id="...", private_key_path="...", config=config)

Pagination

List endpoints return a Page[T] you can iterate, plus a cursor for manual control. For "give me everything" use list_all():

# Manual cursor loop:
page = client.markets.list(status="open", limit=200)
while True:
    for market in page:
        ...
    if not page.has_more:
        break
    page = client.markets.list(status="open", limit=200, cursor=page.cursor)

# Or just:
for market in client.markets.list_all(status="open"):
    ...

# Need a hard cap on pages (e.g. preview / quick sample)?
for market in client.markets.list_all(status="open", max_pages=5):
    ...

*_all() iterates until the server returns no cursor by default. Pass max_pages=N for an explicit bound; passing 0 raises ValueError.

Page[T] also converts to a DataFrame when the optional extras are installed:

pip install 'kalshi-sdk[pandas]'   # or [polars] or [all]

df = client.markets.list(status="open", limit=100).to_dataframe()
# Decimal and datetime preserved as native types in object columns.

Testing against the SDK (no live API)

For SDK consumers who want offline integration tests, kalshi.testing ships record-and-replay transports:

from kalshi import KalshiClient
from kalshi.testing import RecordingTransport, ReplayTransport

# Record once against the real demo API:
with KalshiClient.from_env(transport=RecordingTransport("fixtures")) as c:
    c.exchange.status()

# Replay in tests — no network:
with KalshiClient(transport=ReplayTransport("fixtures")) as c:
    c.exchange.status()  # served from fixtures/GET_*.json

Fixtures are JSON; the fingerprint ignores KALSHI-ACCESS-SIGNATURE and KALSHI-ACCESS-TIMESTAMP so signature drift between record and replay does not break matching. Always .gitignore the fixture directory when recording against an authenticated account — fixtures contain the full response body (balances, positions, PII).

Resources

Documentation site	https://texascoding.github.io/kalshi-python-sdk/
Kalshi REST OpenAPI spec	https://docs.kalshi.com/openapi.yaml
Kalshi WebSocket AsyncAPI spec	https://docs.kalshi.com/asyncapi.yaml
Production base URL	https://api.elections.kalshi.com/trade-api/v2
Demo base URL	https://demo-api.kalshi.co/trade-api/v2
Changelog	CHANGELOG.md
Issues	https://github.com/TexasCoding/kalshi-python-sdk/issues

License

MIT — see LICENSE.