Your agent's safety layer
Your agent can trade. TickTape tells it when not to.
Before your Polymarket agent risks your money, TickTape checks trades against live liquidity and fees, and wallets against historical copyability — and returns OK, CAUTION or VETO.
Polymarket says #1 by volume. TickTape says VETO.
Agent-paid from $0.02 · USDC over x402 · no account · free sandbox
Evidence
18 of 24 active wallets lost money when copied after fees and slippage.
Every bar below is a real wallet, replayed by the same engine that answers your agent's preflight calls.
Chart data as a table (24 wallets)
| Wallet | Fee-inclusive ROI | Copied trades | Win rate |
|---|
Detect → Check → Trade or veto
Your agent finds a trade to place or a wallet it wants to copy.
It asks TickTape — one HTTP call against the live book and the recorded tape, paid in USDC.
OK proceeds. VETO stops the order before it costs you money.
The safety layer
Your agent has data and execution. It's missing the word no.live
Data is a solved problem — PMXT publishes raw feeds free. Execution is a solved problem — PolyCat, Stand, and PMXT route orders. What no one ships is an independent voice that says no before the order goes out. TickTape is that voice, as four paid HTTP tools your agent can call — and pay for — today, plus a prepaid credits endpoint:
agent › → GET ticktape.cc/api/red_list (1 credit)
ticktape › { swisstony: verdict VETO · roi_fee_inclusive −6.2% · 90-day window · first-entry mirror, explicit 2026 taker fees }
agent › No — volume isn't copyability. The leaderboard rates one wallet COPY: S-Works +6.8% fee-inclusive (90d), OOS-validated…
Humans read dated, frozen samples of the same reports — the current feed is what agents buy, one call at a time.
Full agent docs → HTTP + x402 + MCP (https://ticktape.cc/mcp) today · x402 Bazaar listing on the roadmap · machine spec: /llms.txt · /openapi.json · machine manifest: /agent.json
Agents start here — live now
Live calls are paid from the first request — HTTP 402 is the whole onboarding. The 402 carries x402 payment requirements — dual-stack: v2 primary (base64 PAYMENT-REQUIRED response header; retry with PAYMENT-SIGNATURE; network eip155:8453) and v1 accepted (JSON accepts body + X-PAYMENT header). Pay $0.02 in USDC on Base and retry — no account needed. Retrying with the same payment authorization returns the cached response, never double-charges; paid responses include price_paid_usdc and a basescan receipt. Inspect requirements without paying: ?x402_probe=1. Test free forever: ?sandbox=1 on all four tools. Fast lane: POST /api/credits — one $5 payment → 500 credits + bearer token → ~0.4 s calls. Thresholds (max_slippage_c, veto_slippage_c, max_book_age_s) are overridable per call.
| Call | Per-call (x402) | With credits | Status |
|---|---|---|---|
| GET /api/preflight_trade | $0.02 | 1 credit | live |
| GET /api/leaderboard | $0.02 | 1 credit | live |
| GET /api/red_list | $0.02 | 1 credit | live |
| GET /api/preflight_copy | $0.10 | 5 credits | live |
preflight_copy: the only parameter is wallet=0x… (alias user=). Uses a fixed $5-entry / +2c-slippage simulation computed from TickTape's private trade tape — exact size-and-delay replay is not live. Sub-second verdicts for covered wallets, computed from TickTape's own identity-tagged tape running since Dec 2025 — a history Polymarket's live API cannot reproduce. Wallets below the coverage floor return QUEUED free of charge (never charged) — typically ready in minutes, worst case about an hour. COPY verdicts require out-of-sample validation; wallets validated in-sample only are capped at WATCH.
Credits: POST /api/credits — one $5 USDC payment buys 500 credits (1 credit = $0.01 effective) plus a bearer token — half price per call on every tool; credits valid 12 months, the last 5 tokens per wallet stay valid.
Latency, honestly: a per-call x402 payment settles on-chain — expect ~2–5 s added per paid call, longer under congestion. Credit calls answer in ~0.4 s. Sandbox calls ~0.4 s — free, unlimited.
Operator safety — enforce in code, not prompts (v2 @x402/fetch dropped the old default 0.10 USDC cap): max $0.10 per TickTape call; set a daily spend cap; pay only endpoints under https://ticktape.cc/; verify payTo against /llms.txt; USDC on Base only; reuse the same payment authorization on retries.
Your agent is the customer
You give it a small x402 budget. It discovers TickTape, checks the price, pays per verdict and decides whether to proceed. Your only job as the operator:
- Create a separate Base wallet for your agent.
- Put $2 USDC in it — not your trading bankroll.
- Copy the TickTape setup into your agent.
- Enforce a $0.10 per-call ceiling and a $2 daily cap in code.
Receipts
When we say copyable, this is the bar.
Most "edges" die when tested on real data. Every number below comes from the same engine we run on our own money — and survivors must repeat on a second, earlier month before we call them copyable:
| Window | Copied trades | Win rate | ROI (2c slippage basis) |
|---|---|---|---|
| Jun 12 – Jul 11 | 132 | 63% | +11.3% |
| May 12 – Jun 11 (out-of-sample) | 107 | 64% | +2.5% |
+11.3% after the 2c slippage floor; +9.2% after explicit taker fees ($14.02 on 132 trades). Charging fees on all 24 wallets moves the count from 7 positive to 6 — aenews2 slips to −0.9% in-sample (but +17.2% out-of-sample). Full table in the evidence pack.
Of 24 candidate wallets screened the same way, 18 were net-negative to copy after explicit fees (17 on the 2c-slippage basis alone) — including the two biggest whales by volume. The report tells you which is which. See the free sample audit →
Method, honestly: audits price every copied entry at the trader's price plus a conservative 2c slippage floor, $5 per trade, with a drift-based latency sensitivity check — 57% of S-Works entries drift at most 1c within 30–60 s (median drift 1c; we previously published 84% here — our own recompute corrected it). Exact order-book replay at your size and delay is not live — we record the top-200 books, but today's verdicts use the fixed floor above. Full methodology: Tech zone.
For humans
Humans can read. Agents can buy.
A SaaS with no seats, no users and no checkout — only paid agent calls. Everything on this site is free to read: the weekly leaderboard, the sample wallet audit, the methodology and the data documentation. The only way to pay is x402 — your agent (or any x402 client) buys the calls. The plain-language guide, the operator checklist and the wallet-check story live on the human hub.
For humans → · Whole offer → · How the verdicts are computed →
In production. Point your agent at it.
No account or issued API key — agents pay per request with x402, or use an optional prepaid bearer token. HTTP 402 is the whole onboarding. Live calls are paid from the first request in USDC on Base; the sandbox is free and unlimited on all four tools.