Getting started for agentic traders

One Skylit account gives your agent three data sets through one credit balance:

1. Sign in and open the Developer page

Sign in at app.skylit.ai and open Developer. Your API account is created the first time you open it (1 credit = $0.001).

How you get access depends on your membership:

Plans can change. GET /v1/account (free) and the X-RateLimit-Limit header always show the values that apply to you, so have your agent read them at startup.

2. Pick how your agent connects

claude mcp add --transport http skylit https://mcp.skylit.ai/mcp

{
  "mcpServers": {
    "skylit": { "url": "https://mcp.skylit.ai/mcp" }
  }
}

export SKYLIT_API_KEY="paste-your-key-here"

# Claude Code, headless (no browser sign-in)
claude mcp add --transport http skylit https://mcp.skylit.ai/mcp \
  --header "Authorization: Bearer $SKYLIT_API_KEY"

{
  "mcpServers": {
    "skylit": {
      "url": "https://mcp.skylit.ai/mcp",
      "headers": { "Authorization": "Bearer YOUR_API_KEY" }
    }
  }
}

3. Make your first REST calls

Every request sends Authorization: Bearer <key>.

# Free: your balance and limits
curl -s https://api.skylit.ai/v1/account -H "Authorization: Bearer $SKYLIT_API_KEY"

# SPY dealer-positioning levels (king node, gatekeepers, walls)
curl -s "https://api.skylit.ai/v1/gex/levels?symbols=SPY" -H "Authorization: Bearer $SKYLIT_API_KEY"

# Latest scored options flow for SPY
curl -s "https://api.skylit.ai/v1/flow/SPY" -H "Authorization: Bearer $SKYLIT_API_KEY"

# SPY 1-minute bars (from/to are Unix seconds)
curl -s "https://atlas-api.skylit.ai/v1/history?symbol=SPY&resolution=1&from=1790602200&to=1790625600" \
  -H "Authorization: Bearer $SKYLIT_API_KEY"

import os, requests

s = requests.Session()
s.headers["Authorization"] = f"Bearer {os.environ['SKYLIT_API_KEY']}"

levels = s.get("https://api.skylit.ai/v1/gex/levels", params={"symbols": "SPY,QQQ"})
levels.raise_for_status()
for sym in levels.json()["data"]["symbols"]:
    king = sym["kingNode"]  # null when no king node is classified
    print(sym["symbol"], "king", king["strike"] if king else None, "spot", sym["spot"])
print("credits left:", levels.headers.get("X-Credits-Remaining"))

It prints something like this (your numbers will differ):

SPY king 761 spot 765.44
QQQ king 735 spot 736.93
credits left: 4998

GET /v1/account answers with your balance and the limits your agent should respect (abridged; values depend on your plan):

{
  "data": {
    "status": "active",
    "creditsBalance": 4998,
    "balanceUsd": 4.998,
    "limits": {
      "requestsPerMinute": 120,
      "symbolsPerHeatmapCall": 10,
      "symbolsPerStream": 10,
      "streamSymbolsConcurrent": 25,
      "historicalInFlight": 2,
      "activeKeys": 5,
      "streamMaxDurationMinutes": 60
    }
  }
}

Successful responses are {"data": ..., "meta": ...}; errors are {"error": {"code", "message"}}.

4. Try the MCP tools

Ask your agent something like:

Where are SPY's key gamma levels right now, and is today's options flow leaning bullish or bearish?

A good agent calls heat_levels and flow_feed (or underlying_stats), then answers. Useful first tools:

Each tool's description states its credit cost.

See the full tool catalog and example prompts.

5. Watch your credits

• Every chargeable response carries X-Credits-Remaining; MCP results carry creditsRemaining in meta.
• Failed calls are free. Any 4xx or 5xx is refunded.
• Streams (/v1/stream) charge 1 credit per symbol to open, then 1 credit per symbol per minute, including symbols that have no live board yet (outside market hours). The connected event states creditsPerMinute and maxDurationSeconds. A 10-symbol stream left open for an hour costs about 600 credits.
• The Developer page's Usage tab shows spend by endpoint and by key.
• Each endpoint's price is on its API Reference page (x-credits in the OpenAPI specs).

6. Stay inside your plan's limits

Your plan sets your request rate, number of keys, symbols per call, historical calls in flight and streams. GET /v1/account (or the account_usage tool) returns the exact values for your account, so have your agent read them at startup instead of hard-coding them.

Every response carries X-RateLimit-Limit, X-RateLimit-Remaining and X-RateLimit-Reset (Unix seconds) for the key you used; the limit is shared across all Skylit APIs. When X-RateLimit-Remaining reaches 0 you get 429 rate_limited: wait until X-RateLimit-Reset, then retry. Back off with jitter rather than retrying in a tight loop.

If something goes wrong