Bankr

Bankr

Execute crypto trading and DeFi operations using natural language. Two integration options:

1. Bankr CLI (recommended) — Install @bankr/cli for a batteries-included terminal experience
2. REST API — Call https://api.bankr.bot directly from any language or tool

Both use the same API key. The API has two layers:

• Wallet API (/wallet/*) — Direct, synchronous endpoints for portfolio, transfers, signing, and transaction submission
• Agent API (/agent/*) — AI-powered async prompt endpoint for natural language operations

Getting an API Key

Before using either option, you need a Bankr API key. Two ways to get one:

Option A: Headless email login (recommended for agents)

Two-step flow — send OTP, then verify and complete setup. See "First-Time Setup" below for the full guided flow with user preference prompts.

# Step 1 — send OTP to email
bankr login email user@example.com

# Step 2 — verify OTP and generate API key (options based on user preferences)
bankr login email user@example.com --code 123456 --accept-terms --key-name "My Agent" --read-write

This creates a wallet, accepts terms, and generates an API key — no browser needed. Before running step 2, ask the user which APIs they need (wallet, agent, both via --read-write, LLM gateway) and their preferred key name.

Option B: Bankr Terminal

1. Visit bankr.bot/api-keys
2. Sign up / Sign in — Enter your email and the one-time passcode (OTP) sent to it
3. Generate an API key — Create a key with Wallet & Agent API access enabled (the key starts with bk_...)

Both options automatically provision EVM wallets (Base, Ethereum, Polygon, Unichain) and a Solana wallet — no manual wallet setup needed.

Option 1: Bankr CLI (Recommended)

Install

bun install -g @bankr/cli

Or with npm:

npm install -g @bankr/cli

First-Time Setup

Headless email login (recommended for agents)

When the user asks to log in with an email, walk them through this flow:

Step 1 — Send verification code

bankr login email <user-email>

Step 2 — Ask the user for the OTP code and all preferences in a single message. This avoids unnecessary back-and-forth. Ask for:

1. OTP code — the code they received via email
2. Accept Terms of Service (REQUIRED) — Present the Terms of Service link and confirm the user agrees. The login command will fail for new users without --accept-terms. You MUST ask for ToS acceptance and do not pass --accept-terms unless the user has explicitly confirmed.
3. Which APIs do they need?
   • Wallet API — enabled by default, use --no-wallet-api to disable
   • Agent API (--agent-api) — AI-powered prompts and natural language operations
   • Token Launch — enabled by default, use --no-token-launch to disable
   • Add --read-write to allow transactions (without it, enabled APIs are read-only)
4. Enable LLM gateway access? (--llm) — multi-model API at llm.bankr.bot (currently limited to beta testers). Skip if user doesn't need it.
5. Key name? (--key-name) — a display name for the API key (e.g. "My Agent", "Trading Bot")

Step 3 — Construct and run the step 2 command with the user's choices. Do NOT execute if the user has not explicitly accepted the Terms of Service — ask again if needed:

# Full access: wallet + agent with write + LLM
bankr login email <user-email> --code <otp> --accept-terms --key-name "My Agent" --agent-api --read-write --llm

# Agent with write access (AI can execute transactions)
bankr login email <user-email> --code <otp> --accept-terms --key-name "Trading Agent" --agent-api --read-write

# Default key (wallet + token launch, read-only)
bankr login email <user-email> --code <otp> --accept-terms --key-name "My Key"

# Agent read-only (research, prices, balances — no transactions)
bankr login email <user-email> --code <otp> --accept-terms --key-name "Research Agent" --agent-api

# LLM-only (no wallet, no token launch)
bankr login email <user-email> --code <otp> --accept-terms --key-name "LLM Client" --no-wallet-api --no-token-launch --llm

Login options reference

New key defaults (when no flags are passed):

Any option not provided on the command line will be prompted interactively by the CLI, so you can mix headless and interactive as needed.

Login with existing API key

If the user already has an API key:

bankr login --api-key bk_YOUR_KEY_HERE

If they need to create one at the Bankr Terminal:

1. Run bankr login --url — prints the terminal URL
2. Present the URL to the user, ask them to generate a bk_... key
3. Run bankr login --api-key bk_THE_KEY

Separate LLM Gateway Key (Optional)

If your LLM gateway key differs from your API key, pass --llm-key during login or run bankr config set llmKey YOUR_LLM_KEY afterward. When not set, the API key is used for both. See references/llm-gateway.md for full details.

Verify Setup

bankr whoami
bankr wallet portfolio
bankr agent prompt "What is my balance?"

Option 2: REST API (Direct)

No CLI installation required — call the API directly with curl, fetch, or any HTTP client.

Authentication

All requests require an X-API-Key header:

curl -X POST "https://api.bankr.bot/agent/prompt" \
  -H "X-API-Key: bk_YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"prompt": "What is my ETH balance?"}'

Quick Example: Submit → Poll → Complete

# 1. Submit a prompt — returns a job ID
JOB=$(curl -s -X POST "https://api.bankr.bot/agent/prompt" \
  -H "X-API-Key: $BANKR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"prompt": "What is my ETH balance?"}')
JOB_ID=$(echo "$JOB" | jq -r '.jobId')

# 2. Poll until terminal status
while true; do
  RESULT=$(curl -s "https://api.bankr.bot/agent/job/$JOB_ID" \
    -H "X-API-Key: $BANKR_API_KEY")
  STATUS=$(echo "$RESULT" | jq -r '.status')
  [ "$STATUS" = "completed" ] || [ "$STATUS" = "failed" ] || [ "$STATUS" = "cancelled" ] && break
  sleep 2
done

# 3. Read the response
echo "$RESULT" | jq -r '.response'

Conversation Threads

Every prompt response includes a threadId. Pass it back to continue the conversation:

# Start — the response includes a threadId
curl -X POST "https://api.bankr.bot/agent/prompt" \
  -H "X-API-Key: $BANKR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"prompt": "What is the price of ETH?"}'
# → {"jobId": "job_abc", "threadId": "thr_XYZ", ...}

# Continue — pass threadId to maintain context
curl -X POST "https://api.bankr.bot/agent/prompt" \
  -H "X-API-Key: $BANKR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"prompt": "And what about SOL?", "threadId": "thr_XYZ"}'

Omit threadId to start a new conversation. CLI equivalent: bankr agent prompt --continue (reuses last thread) or bankr agent prompt --thread <id>.

API Endpoints Summary

Wallet API (/wallet/*) — Direct endpoints (synchronous)

• Read endpoints (/wallet/me, /wallet/portfolio) — any valid API key with a wallet
• Swap quote (/wallet/swap-quote) — a quote is a read, so read-only API keys are allowed; API-key callers still need walletApiEnabled
• Write endpoints (/wallet/swap, /wallet/transfer, /wallet/sign, /wallet/submit) — require walletApiEnabled and reject read-only keys. /wallet/transfer also enforces allowedRecipients; /wallet/swap does not (output returns to your own wallet)
• IP allowlist enforced on all endpoints

Recipient & user lookup helpers (public, no auth)

The legacy aliases /public/resolve-recipient and /public/search-users have been removed — use the structured /addresses/resolve and /users/search endpoints instead.

Agent API (/agent/*) — AI-powered endpoints (async)

Public endpoints (no auth required)

Removed legacy endpoints

The following /agent/* endpoints have been removed. Use the /wallet/* equivalents:

For full API details (request/response schemas, job states, rich data, polling strategy), see:

Reference: references/api-workflow.md | references/sign-submit-api.md

CLI Command Reference (v0.3.x)

@bankr/cli 0.2+ organizes commands into three namespaces: wallet, agent, and tokens. Old flat commands (balances, prompt, status, etc.) still work as deprecated aliases.

bankr wallet — Wallet Operations

bankr agent — AI Agent Operations

bankr tokens — Token Discovery

bankr files — File Storage

bankr club — Bankr Club Membership

Manage Bankr Club subscription from the CLI (status, signup, cancel). Pay with USDC (default), BNKR, ETH, or any Base ERC-20 — non-USDC/BNKR tokens are swapped to USDC at checkout.

Pricing is $20/mo or $198/yr USD-equivalent. Actual on-chain amount depends on the chosen token's price at quote time. The --token flag was added in CLI 0.3.4; older versions only support USDC.

What Club actually gates (everything else works without it):

The launch cap is the one perk row that is deliberately flat: every Bankr wallet — Standard, Club, partner organization, provisioned partner — gets the same 3 counted launch attempts per rolling 24 hours. See Token Deployment.

Max Mode is the alternative to a subscription — pay per request from LLM credits for unlimited terminal messages, and it works with external/connected wallets, which Club checkout does not (Club requires an embedded Bankr wallet). On the Agent API, non-Club Max Mode is still capped at 100 requests/day. Holding a legacy Bankr Club NFT does not grant membership; it was a commemorative token for the first 1,000 subscribers.

Monthly → yearly upgrade: active monthly members can upgrade to yearly by asking the agent (e.g. "upgrade me to yearly"). The full yearly price is charged and the new 365-day term stacks on top of any remaining monthly time, so no paid time is lost. Yearly → monthly downgrades and yearly resubscribes are not supported.

Auth & Config Commands

Valid config keys: apiKey, apiUrl, llmKey, llmUrl

Default config location: ~/.bankr/config.json. Override with --config or BANKR_CONFIG env var.

Non-Interactive Mode

For headless environments — CI pipelines, Docker containers, cron jobs — pass the --not-interactive (alias --ni) global flag, or set BANKR_NOT_INTERACTIVE=1. Both forms work before or after the subcommand.

The flag implies --yes on every confirmation prompt and fails fast (exit 1, clear error) when a command would otherwise hang on a required prompt. Use it whenever the user asks for an automated/scripted invocation.

Read-only commands (whoami, wallet portfolio, tokens, agent status, etc.) don't prompt at all — --ni is harmless but redundant.

Examples:

# Cron — claim creator fees nightly
BANKR_PRIVATE_KEY=0x... bankr --ni fees claim-wallet --all

# CI — programmatic token launch (simulate, no broadcast)
bankr --ni launch --name MyToken --symbol MTK --simulate

# Pipeline — agent prompt from stdin
echo "summarize today's trades" | bankr --ni agent prompt

Deprecated Aliases

Old flat commands still work but prefer the namespaced versions:

Environment Variables

Environment variables override config file values. Config file values override defaults.

LLM Gateway Commands

On the launcher commands (bankr claude, bankr llm claude, bankr llm opencode) all arguments — including -h / --help — are forwarded to the spawned tool, and a help run doesn't require authentication. Use bankr --help / bankr llm --help for the Bankr CLI's own help. Named commands also take precedence over the prompt fallthrough, so bankr claude ... launches Claude Code instead of prompting the agent; use bankr agent "..." when a prompt starts with a command name.

The bankr claude alias and the --help passthrough require @bankr/cli 0.3.18+; on older versions use bankr llm claude.

Core Usage

Simple Query

For straightforward requests that complete quickly:

bankr agent prompt "What is my ETH balance?"
bankr agent prompt "What's the price of Bitcoin?"

The CLI handles the full submit-poll-complete workflow automatically. You can also use the shorthand — any unrecognized command is treated as a prompt:

bankr What is the price of ETH?

Interactive Prompt

For prompts containing $ or special characters that the shell would expand:

# Interactive mode — no shell expansion issues
bankr agent prompt
# Then type: Buy $50 of ETH on Base

# Or pipe input
echo 'Buy $50 of ETH on Base' | bankr agent prompt

Conversation Threads

Continue a multi-turn conversation with the agent:

# First prompt — starts a new thread automatically
bankr agent prompt "What is the price of ETH?"
# → Thread: thr_ABC123

# Continue the conversation (agent remembers the ETH context)
bankr agent prompt --continue "And what about BTC?"
bankr agent prompt -c "Compare them"

# Resume any thread by ID
bankr agent prompt --thread thr_ABC123 "Show me ETH chart"

Thread IDs are automatically saved to config after each prompt. The --continue / -c flag reuses the last thread.

Manual Job Control

For advanced use or long-running operations:

# Submit and get job ID
bankr agent prompt "Buy $100 of ETH"
# → Job submitted: job_abc123

# Check status of a specific job
bankr agent status job_abc123

# Cancel if needed
bankr agent cancel job_abc123

LLM Gateway

The Bankr LLM Gateway is a unified API for Claude, Gemini, GPT, Grok, DeepSeek, Qwen, Kimi, MiniMax, GLM, and other models — multi-provider access, cost tracking, automatic failover, and SDK compatibility through a single endpoint.

Base URL: https://llm.bankr.bot | Dashboard: bankr.bot/llm | API Keys: bankr.bot/api-keys

Key Concepts

• Uses your llmKey if configured, otherwise falls back to your API key
• LLM credits (USD) and trading wallet (crypto) are completely separate balances — having crypto does NOT give you LLM credits
• New accounts start with $0 LLM credits — top up via bankr llm credits add 25 or at bankr.bot/llm?tab=credits before making any LLM calls, or you will get a 402 error
• Check credits: bankr llm credits | Top up: bankr llm credits add <amount> | Auto top-up: bankr llm credits auto --enable --amount 25 --tokens USDC
• In OpenClaw config, prefix model IDs with bankr/ (e.g. bankr/claude-sonnet-5). In direct API calls, use bare IDs (e.g. claude-sonnet-5). Run bankr llm models for the current model list
• Claude Code's [1m] context-tier suffix is optional through the gateway — it's stripped before model lookup and the 1M window comes from the model's own context window, so claude-opus-5 and claude-opus-5[1m] behave identically. If you do use it, quote it (--model "claude-opus-5[1m]") — in zsh it's a glob character class and the command aborts before the CLI runs
• Per-model discounts available for Bankr Club members and partners — applied automatically at billing time
• Image generation: generate images via the OpenAI-native /v1/images/generations endpoint (model gpt-image-2), billed from the same LLM credit balance — see the reference
• Expiring credit grants: promotional or developer grants may carry an expiry date. Your spendable balance is your permanent (purchased) credits plus any unexpired grants — grants are spent first (soonest-expiring first) and drop off automatically at expiry
• Privacy tiers: every request is served at standard, zdr (zero data retention), or private (TEE). Ask for a tier per request, per model, per base URL, or account-wide — see below

Privacy Tiers

Three nesting levels of data-handling guarantee, requested the same way on every endpoint:

Four ways to ask, so any client can reach any tier:

# 1. The `privacy` request field
curl -X POST https://llm.bankr.bot/v1/chat/completions -H "X-API-Key: $BANKR_LLM_KEY" \
  -d '{"model":"glm-5.2","privacy":"zdr","messages":[{"role":"user","content":"Hello"}]}'

# 2. A base-path prefix — for tools that only take a base URL, key, and model
export OPENAI_BASE_URL=https://llm.bankr.bot/zdr/v1      # OpenAI-compatible clients
export ANTHROPIC_BASE_URL=https://llm.bankr.bot/zdr      # Anthropic-compatible clients

# 3. A model-ID suffix
bankr llm models --zdr                                   # which models have a ZDR slot
# → use `glm-5.2:zdr` or `glm-5.2:private` as the model ID

# 4. Account-wide — Settings in the web terminal, applies to every request

• Every tier fails closed. If no provider can serve your model at the tier you asked for, the request is rejected (422 zdr_unavailable) — never quietly downgraded.
• Account setting and request combine, strongest wins. A request can tighten past the account setting; nothing in a request can drop below it.
• A tier endpoint is authoritative. Sending a request that names a different tier to /zdr or /private returns 400 privacy_conflict rather than merging. An unparseable privacy value returns 400 invalid_privacy rather than being ignored.
• X-Privacy-Tier response header reports the tier the request was actually handled under, so the guarantee is verifiable rather than assumed.
• Tier matching is case-insensitive, and only a trailing :zdr / :private token counts as the suffix opt-in.

Max Mode — Pick the Agent's Model

Max Mode overrides the Bankr agent's default model (currently Gemini 3.8 Flash) with any model from the gateway, billed per token from your LLM credit balance (not your trading wallet):

bankr agent "analyze my portfolio" --model claude-opus-5     # or -m
bankr agent "what are the top memecoins today?" -m gemini-3.1-pro
bankr agent prompt "tell me more" --continue --model claude-sonnet-5

The setting is stored on your wallet and applies across every surface — CLI, web terminal, social platforms, and automations. In the web terminal, toggle the Max button and pick a model from the picker.

• Credits are enforced per LLM call, not after the fact. Your spendable balance (minus anything already metered but not yet deducted) becomes the run's budget, re-checked before every call — the turn ends honestly when it's exhausted instead of overspending.
• Shortfalls survive. If a batch can't be fully covered, the credit is left untouched and the usage stays owed, so a later top-up settles it rather than the overspend being written off.
• Top up before enabling, or Max Mode messages will fail.

Quick Commands

bankr llm models                           # List available models
bankr llm credits                          # Check credit balance
bankr llm credits add 25                   # Top up $25 credits (defaults to Base USDC)
bankr llm credits add 25 --token USDT      # Pay USDT on whichever chain holds the most
bankr llm credits add 25 --token ETH       # Native token; auto-swapped on its chain
bankr llm credits auto --enable --amount 25 --tokens USDC,USDT  # Multi-chain auto top-up
bankr llm setup openclaw --install         # Install Bankr provider into OpenClaw
bankr llm setup claude                     # Print Claude Code env vars
bankr claude                               # Launch Claude Code through gateway
bankr claude --model claude-opus-5         # Any tool flags are forwarded
bankr llm opencode                         # Launch OpenCode through gateway

Agent Credit Top-Up

The AI agent can top up your LLM credits directly in conversation — no CLI or web dashboard needed:

bankr agent prompt "Top up my LLM credits with $25"
bankr agent prompt "Add $10 of LLM credits using my ETH"
bankr agent prompt "How many LLM credits do I have left?"

The agent can also report your current LLM credit balance in conversation — including any expiring grants and when they lapse — without needing the bankr llm credits command.

Sending credits to another Bankr user. Credit is transferable peer-to-peer — ask the agent ("send $20 of LLM credits to @alice") or POST /llm/credits/transfer. The recipient must already be a Bankr user (X username or 0x address; no ENS on this path), the minimum is $1, and a wallet may send at most $500 per trailing 24 hours. Read-only API keys are refused; pass your own transferId so a retry can't double-send.

Granted credit can be sent too, behind an opt-in. By default only purchased credit moves. Set useGrantedCredits (the transfer_llm_credits tool flag, or the "use granted credits" toggle in the web Send Credits panel) to let a transfer also spend granted credit — operator grants and bonuses, including ones carrying an expiry:

• Purchased credit still spends first, and free. Only the overflow into the granted slice carries a 10% burn fee, charged on top of the amount.
• The recipient always receives exactly the amount you entered. You are debited amount + fee, and the fee is burned — it isn't credited to anyone.
• Each granted dollar is fee'd once. Concurrent sends can't double-claim the same slice, and a replayed transferId reports the original fee rather than charging again.
• Leaving the flag off keeps the old behaviour byte-for-byte, so existing integrations need no change.

Two different ceilings, and the smaller one binds. Your balance limits what you own; your daily spend budget clamps what can move today. GET /llm/credits/state is the canonical read for both — creditBalanceUsd (purchased pool), creditGrantsUsd and creditGrants[] (live grants with their expiries), totalCreditsUsd, and a dailyBudget object (limitUsd, spentUsd, remainingUsd; limitUsd: null means uncapped). So a wallet holding $265 that's temporarily gated to $10 by a rolling window reads as exactly that, rather than as if the money were missing. The agent quotes both figures and names which one is binding when you ask about your balance.

Endpoint split: GET /llm/usage is usage-only — token counts and spend over a historical window. It no longer carries credit-balance or transferable fields; read credit state from GET /llm/credits/state. Over the daily transfer cap, POST /llm/credits/transfer returns 429.

Daily spend budget. You can cap what the gateway spends on your behalf, independently of your balance. The window is a trailing 24 hours, not a calendar day — nothing resets at midnight, capacity returns as charges age out. Over budget, spending requests return 402 with type: daily_budget_exceeded (distinct from insufficient_credits), while read-only GET endpoints keep working so you can poll for when you're unblocked. The same budget also ends Max Mode agent runs early.

1 credit = $1 USD. Multi-chain: pay with USDC or USDT directly on Base, Polygon, Ethereum, Arbitrum, or BNB Chain, or with any other ERC-20 (auto-swapped to the chain's preferred stablecoin — USDC on most chains, USDT on BNB). When using --token, the CLI picks the chain with the highest USD balance of that token. Maximum $1,000 per top-up.

Model Deprecation

The gateway supports model deprecation with auto-redirect to replacement models. Deprecated models return X-Model-Deprecated and X-Model-Replacement response headers. Hard-deprecated models return HTTP 410 — update your model ID to the replacement indicated in the header.

For full details — setup paths, model list, provider config, SDK examples, key management, and troubleshooting — see:

Reference: references/llm-gateway.md

Capabilities Overview

Trading Operations

• Token Swaps: Buy/sell/swap tokens across chains
• Cross-Chain: Bridge tokens between chains
• Limit Orders: Execute at target prices
• Stop Loss: Automatic sell protection
• DCA: Dollar-cost averaging strategies
• TWAP: Time-weighted average pricing

Reference: references/token-trading.md

Portfolio Management

• Check balances across all chains (bankr wallet portfolio or GET /wallet/portfolio)
• View USD valuations with optional PnL tracking (--pnl or ?include=pnl)
• View NFT holdings (--nfts or ?include=nfts)
• Track holdings by token or chain
• Real-time price updates
• Multi-chain aggregation
• Wrapping/unwrapping the native token (ETH ↔ WETH and equivalents) is reflected in both balances right away
• Filter by chain: bankr wallet portfolio --chain base,solana or GET /wallet/portfolio?chains=base,solana

Reference: references/portfolio.md

Market Research

• Token prices and market data
• Technical analysis (RSI, MACD, etc.)
• Social sentiment analysis
• Price charts
• Trending tokens
• Token comparisons
• Holder snapshots — largest-first holder lists with USD value and supply percentage, plus a concentration summary. Supports a minimum-USD floor, which makes it usable for airdrop targeting ("holders of X with $50+")

Reference: references/market-research.md

Transfers

• Send to 0x addresses, ENS-style names (.eth, .base.eth, .cb.id), or social handles
• CLI direct (bankr wallet transfer) accepts 0x addresses + ENS only — social handles go through the AI agent
• Multi-chain support
• Bulk / multi-recipient sends via the agent — same-chain ERC-20 transfers to many recipients batch into a single on-chain transaction (one set of gas)
• Flexible amount formats
• Social handle resolution (Twitter, Farcaster, Telegram) via the agent

Reference: references/transfers.md

NFT Operations

• Browse and search collections
• View floor prices and listings
• Purchase NFTs via OpenSea — including listings priced in an ERC-20 rather than the native token (e.g. USDG on Robinhood Chain). The token approval, the payment-currency balance check, and decimals-aware pricing are all handled for you
• Floor buys survive a snipe. Asking for "the cheapest X" without naming a token ID now walks the candidate listings: if the one Bankr picked is filled by someone else mid-purchase (or the order goes invalid, or the signer's simulation reverts), it moves to the next-cheapest listing instead of failing the whole request — up to three purchase attempts, and the reply names the substitution so it's never silent. Naming an explicit token ID stays a single attempt, by design
• Accept offers on NFTs you own
• View your NFT portfolio
• Transfer NFTs
• Mint from supported platforms

Reference: references/nft-operations.md

Onchain Reward Campaigns (Merkl)

The agent can read and claim Merkl incentive rewards — the campaign rewards protocols distribute to liquidity providers and holders — on Base and Robinhood Chain:

• Check what a wallet has earned and what's currently claimable, per chain, with USD totals. Only your own wallet address is accepted
• Claim the claimable rewards on one chain in a single transaction. Claiming signs from your embedded Bankr wallet, so it isn't available in wallet mode (external/connected wallets)
• Discover earn opportunities — live Merkl campaigns ranked by APR, with TVL, daily rewards, protocol and deposit link; optionally filtered to a token symbol

bankr agent prompt "Do I have any Merkl rewards to claim?"
bankr agent prompt "Claim my Merkl rewards on base"
bankr agent prompt "What are the best Merkl earn opportunities for USDC on base?"

Campaign availability and eligibility change over time — ask the agent for the live picture rather than assuming a campaign is still running.

Polymarket Betting

• Search prediction markets
• Check odds
• Place bets on outcomes
• View positions — including any unspent collateral left in the Polymarket deposit wallet, which is recoverable with "sweep my polymarket deposit wallet"
• Redeem winnings

Reference: references/polymarket.md

Leverage Trading

• Hyperliquid (primary) — Perpetual futures on Hyperliquid L1 with on-chain order book. Crypto, stocks (TSLA, AAPL, NVDA via HIP-3), spot trading. Up to 50x leverage.
• Avantis (secondary) — Perpetuals on Base for crypto, equities (NVDA, TSLA, AAPL, HOOD, and more), forex, and

…(truncated)