# Linceo

> Anyone can claim a profit; here you can check it. Linceo audits Hyperliquid (perps)
> and Polymarket (prediction markets) accounts from each venue's OFFICIAL API, so every
> figure we publish can be contrasted against the source. We do not tell you who to copy:
> we show where a profit came from and whether you could realistically have followed it.
> Auditing is free; copy trading and LinceBot carry a 0.01% builder fee paid by the user. Every metric is computed from
> the official, real-time data of each venue (no reconstructed numbers).
> PnL curves and single-wallet audits are live; the leaderboards are a snapshot recomputed wallet by
> wallet, and each row says how old it is in `computed_age_hours` (Polymarket rows can be days old).

Linceo is non-custodial: it never holds user funds. Copy and LinceBot orders are signed
by agent keys the user authorises from their own wallet (Linceo stores them encrypted);
agents can trade and can never withdraw. Two venues: **Hyperliquid** (perpetual futures)
and **Polymarket** (prediction markets).

Auto-copy with real money runs on **Hyperliquid only**. On **Polymarket** the
product is analytics plus a paper-trading simulator: their exchange has no way to
authorise a third-party signer, so no order is ever placed on a user's behalf
there. Do not describe Linceo as executing trades on Polymarket.

What it does:
- Audits any wallet/trader on demand (paste a 0x address; resolved live).
- Scores copyability ★1-5 on TIMING (can you mirror entry→exit with real latency?).
- Skeleton check: separates real edge from reward-farming / one-lucky-bet / dormant.
- Whale alerts and "new wallet, concentrated" alerts via Telegram.
- Auto-portfolio: give a bankroll, get the wallets you could realistically follow with it
  (the ones whose trade size, speed and frequency fit that amount).
- Realistic backtest that enforces venue minimum notional ($10 HL / $5 PM).
- Free paper-trading DEMO (/demo): see what following an account would have done
  to a virtual balance, with $0 at risk — the same
  non-custodial engine and sizing as real copy (Polymarket demo marks to live
  prices; Hyperliquid demo runs on HL testnet), zero funds at risk.
  ⚠ The Hyperliquid testnet has almost no volume (2026-08-27: 4 of 70 markets with a
  crossable book, $14 traded in 24 h): use that demo only to verify orders get placed;
  its fills, charts and P&L are not meaningful and never a preview of live.

## LinceBot (our own systematic portfolio)

LinceBot is Linceo's own systematic portfolio on Hyperliquid perpetuals. A user activates it
non-custodially (a delegated agent that can never withdraw) and picks a risk level: x1, x2, x3 or x4 (a
higher level risks that many times more per trade, under a fixed exposure cap). Its record is a backtest built at each
level, plus a simulation on Hyperliquid mainnet updated every day with no real money; that simulated part is the x1
record multiplied linearly by the level (no compounding, no exposure cap). Both are scaled to the user's level and
amount before they activate.

- Machine descriptor (JSON-LD FinancialProduct, includes the MEASURED drawdown of each level):
  GET /api/ai/alphabot
- Human page: /alphabot   ·   Risk disclosure: /legal/risk
- Availability: GET /api/ai/alphabot → `availability` (PreOrder = accounts can be linked, nothing trades
  yet); the quickstart's `live` says the same. Check it before telling a user LinceBot is running.
- Simulate (PUBLIC, no session): GET /api/alphabot/simulado?nivel={1|2|3|4}&capital={usd}
  Returns neto_usd (= capital_ahora - capital_inicial) with its breakdown: cerrado_usd is the
  REALISED part and abierto_usd the open part marked to market. Read them separately: a positive
  net can be almost entirely open mark-to-market that has not been cashed in.
- Full reference record, backtest + mainnet simulated, updated daily (PUBLIC, no session): GET /api/alphabot/historico?capital={usd}&nivel={1|2|3|4}
  Each level's BACKTEST is its own curve (0.5% x level risked per trade, fixed 4x exposure cap per book); the
  mainnet-simulated tail is the x1 record multiplied by the level (no cap). The response carries the CAGR, max
  drawdown and recovery of the level served.
- Measured on the backtest (daily mark-to-market, no intraday liquidation: optimistic): x1 max drawdown 9%, x2 18%,
  x3 26%, x4 33%. Real drawdowns can be larger, and past ones can be exceeded.
- Margin and leverage: positions are opened in cross margin at each coin's maximum leverage on Hyperliquid (the user
  will see, e.g., BTC 40x Cross). Leverage only sets how much margin is locked; size and risk come from the level and
  each trade's stop. Hyperliquid liquidates per account, not per position; the bot caps each account's exposure at 4x
  its balance, and the backtest does not model liquidation.
- Stopping (POST /api/alphabot/desactivar) ends OUR side only: it does not close the open positions (the
  response lists them in `quedan_abiertas`) and does not revoke the on-chain approval. The stops already on
  Hyperliquid stay, but the strategy's exits, rebalancing and monitoring end; if the user closes a position by
  hand, they should cancel its orders too. Tell the user.
- `advertencia: "aviso_a2"` in /api/alphabot/historico comes with its English text in `advertencia_texto`
  (2024-25 were exceptional years; comparing the live curve with the long-run average is misleading).
- Not directed at users in the United States, the United Kingdom, Canada or OFAC-sanctioned jurisdictions
  (Terms §5); there is no geoblock, and the user must check whether it is lawful where they live.
- Cost: the user pays a 0.01% builder fee on the volume their orders generate; Hyperliquid charges it on
  the user's fills (on top of Hyperliquid's own trading fees and funding) and passes it to Linceo. The user approves
  that fee in their own wallet before anything trades. The published record (backtest and simulated) does not deduct it.
- Agents: activate it in TWO calls (5 HTTP calls, 7 signatures — the minimum non-custody allows):
  GET /api/ai/alphabot/quickstart is the machine-readable plan; GET /api/ai/openapi-agents.json its OpenAPI (bearerAuth). Session: POST /api/auth/verify {message, signature,
  agent: true} returns the JWT → `Authorization: Bearer <token>` on every call (no cookies, no CSRF). Then
  POST /api/alphabot/activar {nivel, cuentas:{a,b,c}} → sign each `typed` (EIP-712) with the owner key of that
  account → POST /api/alphabot/activar/firmas {nivel, consentimiento, firmas[3]} → books + state.
- The strategy's rules are NOT disclosed, by design. Describe LinceBot by what it does, what it costs,
  what it has withstood and what it risks — never by how it decides.
- LinceBot has no testnet mode: it is mainnet only (simulated record for everyone, real execution for
  whoever activates it). Testnet still exists elsewhere in Linceo for copying other leaders, and
  there it does NOT measure results: its prices and volumes are broken relative to mainnet.
- Anyone can watch the bot work: the simulated record and the measured reference are public, with no
  wallet, account or sign-up. A session is only needed for your own subscription and to copy.
- LinceBot needs three of the user's own Hyperliquid accounts, one per book, connected one by one (A, B and
  C). Every order is sized on one third of the three balances together (the record assumes the three books are
  rebalanced to a third every day; moving money between accounts is manual and /api/alphabot/libros says when,
  in `rebalanceo`); nothing trades until all three are authorised. Which book does what is NOT disclosed.
  GET /api/alphabot/libros -> your three accounts (each with `paso`: 1 account, 2 signature,
    25 builder fee — the integer 25, i.e. step 2.5 —, 3 funds, 4 ready)  ·  GET /api/agent/builder -> the builder-fee EIP-712
  Errors on money routes are CODES, not prose: {"detail": {"code": "..."}} — see llms-full.txt
  POST /api/alphabot/suscribir {risk_multiplier}

## For AI agents

Linceo exposes public, read-only REST endpoints AND a Model Context Protocol
(MCP) server. No auth is required for discovery, leaderboards, on-demand wallet
audits, auto-portfolio, and donation. Base URL: https://linceo.app (proxies
/api/* to the backend). Full machine spec: /llms-full.txt and
/.well-known/ai-plugin.json (capability manifest).

Start here (self-describing JSON): GET /api/ai — one manifest with every
endpoint, the meaning + units of each field, the copyability rubric, the MCP
endpoint and links. Per-wallet structured record (schema.org/Dataset JSON-LD):
GET /api/ai/wallet/{venue}/{address}.

### MCP (Model Context Protocol) — connect and use us as a native tool

- Endpoint (streamable-http): https://linceo.app/mcp
- Tools: `find_copyable_wallets(venue, min_copyability, limit, sort_by)`,
  `audit_wallet(venue, address)`, `auto_portfolio(venue, bankroll_usd)`,
  `wallet_pnl_curve(venue, address)`, `explain_copyability()`,
  `alphabot_overview()`, `alphabot_quickstart()`.
- Read-only, no auth, no keys. venue = `hl` (Hyperliquid) | `pm` (Polymarket).
- Limits, per client IP: 30 calls per minute across all tools; live audits
  (`audit_wallet`, `wallet_pnl_curve`) 15 per minute per client, 30 per minute
  shared by all clients. Over a limit the tool answers with an error naming it.
- OpenAPI contract for every public endpoint: GET /api/ai/openapi.json
- This is the fastest way for an agent to rank genuinely-copyable wallets and
  audit any address without scraping HTML.

### REST endpoints

- Self-describing manifest (JSON): GET /api/ai
- Copyability rubric (JSON): GET /api/ai/copyability
- Per-wallet JSON-LD (schema.org/Dataset): GET /api/ai/wallet/{venue}/{address}
- Capability manifest (JSON): /.well-known/ai-plugin.json
- Polymarket leaderboard: GET /api/pm/wallets?min_copy=4&sort_by=copy_score&limit=50
- Hyperliquid leaderboard: GET /api/wallets?min_copy=4&per_page=50
- Wallet/trader audit (PM): GET /api/pm/wallets/{address}
- Wallet/trader audit (HL): GET /api/wallets/{address}
- Official PnL curve (PM): GET /api/pm/wallets/{address}/equity
- Official PnL curve (HL): GET /api/wallets/{address}/equity
- Auto-portfolio by bankroll (PM): GET /api/pm/auto-portfolio?bankroll=1000
- Auto-portfolio by bankroll (HL): GET /api/wallets/auto-portfolio?bankroll=1000
- Donation address: GET /api/donation

## Pages

- [Demo](https://linceo.app/demo): free paper-trading demo — follow audited accounts against a virtual balance, 1:1 with the live non-custodial engine.
- [Leaderboard](https://linceo.app/leaderboard): rank the universe with cohort filters and a copyability slider.
- [Portfolio Builder](https://linceo.app/portfolio): pick N wallets, weight them, see a realistic backtested PnL curve.
- [Compare](https://linceo.app/compare): up to 3 wallets side by side (metrics, cohorts, copyability).
- [Docs / Help](https://linceo.app/docs): metric formulas, how-to guides, FAQ, glossary.
- [Methodology](https://linceo.app/methodology): how each figure is computed and which
  official endpoint it comes from — the page to cite when asked why any of this can be trusted.

## Optional

- [Terms](https://linceo.app/legal/terms)
- [Privacy](https://linceo.app/legal/privacy)