Context runtime for your
prediction market agents

Store a core thesis. Accumulate context — injected by you or scanned from the web. The runtime monitors markets 24/7, finds alpha, and pushes it to your agent. CLI and API.

npm install -g @spfunctions/cli

Free plan · charge by token after 15M

Prebuilt context packages

Elections, geopolitics, economics — pre-structured context your agent can load instantly. No cold start.

24/7 market scan

Runtime checks Kalshi and Polymarket against your thesis continuously. Surfaces mispriced contracts. Push or pull.

Agent-native API

One call to get context, get alpha, inject signals. Built for trading bots and agents — not dashboards.

sf agent — terminal
$ sf agent f582bf76
SF Agent — f582bf76  82%  claude-sonnet-4.6

> show my P&L
⚡ get_positions ✓ (0.4s)

  KXWTIMAX-T135   YES  795 @ 57¢  now 60¢  +$23.85  ▲
  KXRECSSNBER-26  YES  3022 @ 32¢  now 35¢  +$90.66  ▲
  Total P&L: +$114.51

> which contracts have good liquidity to enter?
⚡ get_context ✓ (0.3s)

  WTI $150  38¢  edge +36  spread 1¢  high  ★
  Gas $4.50  14¢  edge +41  spread 2¢  high  ★
  WTI $140  50¢  edge +29  spread 2¢  high

> note: Brent just broke $100
⚡ inject_signal ✓ (0.2s)
✓ Signal queued for next evaluation
works with any agent framework
Claude Code
> check my thesis for recession trades

Tool: sf context f582bf76 --json
Recession 2026 YES @ 35¢ — your thesis implies
55¢. Edge: +20¢, liquidity high.
Recommend: increase position.
OpenClaw
claw.add_tool("sf_context", endpoint="/api/thesis/:id/context")
claw.add_tool("sf_signal", endpoint="/api/thesis/:id/signal")

# agent auto-calls sf_context every cycle
# injects observations via sf_signal
→ "WTI $135 moved +8¢, triggering eval..."
Any agent via CLI
sf context f582bf76 --json  # read state
sf signal f582bf76 "..."    # write signal
sf evaluate f582bf76        # trigger eval
sf agent f582bf76           # interactive

Left: interactive CLI agent. Right: plug into Claude Code, OpenClaw, or your own agent. Same runtime, same API.