Skip to content

SpectralEdge developer docs

SpectralEdge reads Kalshi’s 15-minute markets on Bitcoin (KXBTC15M), Ethereum (KXETH15M), Solana (KXSOL15M) and XRP (KXXRP15M) and raises signals: a side (up or down), a calibrated confidence, the venue’s entry price and an edge estimate. When the market settles, every signal is graded against the result. You can reach the same signals the web app shows three ways.

MCP server

Ask Claude, Cursor or any MCP client about the live contest and settled history. Sign in with your Cotes account.

https://mcp.cotes.ai/mcp

Connect an assistant →

Open-source examples

MIT-licensed stream clients in Python and TypeScript, Discord alerts, a paper-trade logger and a Claude Code plugin with five analysis skills.

Browse the repo →

REST + SSE API

HTTPS JSON for live and settled contests, and a Server-Sent Events stream of openings, signals and settlements.

Quick start ↓

Every route needs a Cotes account with a live trial or subscription. The API and MCP publish signals only; they never place orders.

Terminal window
claude mcp add --transport http spectraledge https://mcp.cotes.ai/mcp

Then run /mcp in Claude Code and sign in. Setup for claude.ai, Cursor and other clients is on the MCP server page.

Terminal window
export COTES_API_KEY='your-key'
export BASE='https://api.cotes.ai'
# Liveness (no auth)
curl -sS "$BASE/v1/health"
# Active contest for a series
curl -sS -H "Authorization: Bearer $COTES_API_KEY" \
"$BASE/v1/contests/active?series=KXBTC15M"

There’s no first-party SDK: call HTTPS and JSON from any language, or start from the examples.

Base URL: https://api.cotes.ai

Method Path Auth
GET /v1/health none
GET /v1/series API key
GET /v1/contests/active?series={series} API key
GET /v1/contests?series={series} API key
GET /v1/contests/{ticker} API key
GET /v1/stream/contests?series={series} API key (SSE)

series is a series prefix: KXBTC15M, KXETH15M, KXSOL15M or KXXRP15M. /v1/series lists them. Use the full ticker only on /v1/contests/{ticker}.

  • Dashboards and alerts: stream new signals as they’re raised, and act on SpectralEdge’s own calls (fired) or filter on your own confidence or edge thresholds.
  • Research: pull settled history, with every signal already graded, to measure how the signals perform under your own rules — or ask an assistant to do it through the MCP server.
  • Bots and agents: feed the signal into your own decision logic.