Katoshi vs Senpi: automation you code, or an AI agent that writes and runs the strategy for you
Katoshi and Senpi both automate trading on Hyperliquid, and that is where the overlap ends. Katoshi is an automation engine you program: TradingView alerts, webhooks and its own trigger language fire orders through a 19-endpoint Signal API, its Agents run on a language model you pay for with your own OpenRouter key, and a new beta app runs Rust strategies with backtests and paper trading on a dedicated Tokyo server. Senpi is an AI trading agent: you describe a strategy in a sentence and the agent researches the market, writes the rules, runs them 24/7 under a risk harness and protects every strategy built from a rule or a template with DSL, from the Senpi app, from Claude or ChatGPT, or from the Chrome Companion.
The honest split: Senpi wins where you want the strategy written, run and protected for you and a scored read of what proven wallets are doing; Katoshi wins where you want to code the logic yourself, test it on history and paper first, and wire TradingView or your own scripts straight into Hyperliquid. Facts below come from katoshi.ai, beta.katoshi.ai and Katoshi’s docs (bots, agents, groups, fees and credits, MCP), DefiLlama’s on-chain record of Katoshi’s builder fees, senpi.ai/pricing and senpi.ai/fees, as of Nov 9, 2026.
TL;DR
| Criterion | Katoshi | Senpi | Winner |
|---|---|---|---|
| Best for | Systematic traders who write their own rules or code, or send TradingView alerts, and want them executed on Hyperliquid | Traders who want an agent to build, run and protect Hyperliquid perp strategies, reachable from the AI tools and sites they already use | - |
| AI agent and connectors | Katoshi Agents: a chat, API or webhook agent on a model you choose and pay for through your own OpenRouter key, with trading tools; an MCP server for Claude Desktop or n8n | Your own Senpi agent inside Claude, ChatGPT, Grok, Hermes and Muse through Senpi Connectors: research, build, run, review or close strategies without opening the app; no model key to bring | Senpi |
| Building a strategy without code | Trigger rules in Katoshi’s expression language (for example ta.crossunder(ta.rsi($$, 14), 30)) or Rust code in the beta app; or alerts from TradingView | Describe the thesis in a sentence; the agent writes the rules, shows them to you, then runs them 24/7 on a hosted runtime under a deterministic risk harness; edits apply live | Senpi |
| Senpi Signals vs Katoshi’s market read | Technical indicators (RSI, MACD, Bollinger, EMA, ATR, VWAP and more) inside triggers and agent chats; no read of what other wallets are doing | Senpi Signals: per asset, the biggest recent moves from wallets with a proven record set against the crowd, scored, with the driver named; Quant Desk scores any Hyperliquid address | Senpi |
| Senpi Companion on the sites you read | None; Katoshi runs in its web app | A Chrome extension on 28 sites (X, Hyperliquid, TradingView, Binance, Coinbase, OKX and more): Quant Desk on any wallet you see, the Signals brief, and Trade the Take, which turns a post into a trade setup | Senpi |
| Protection and exits | Stop-loss and take-profit per signal, several partial take-profits, a maximum order size, a coin whitelist and an on/off switch per bot; no trailing stop or account-level halt in the docs | DSL, a stop that ratchets up as a trade gains, on every strategy built from a rule or a template (copy strategies carry at most a flat stop); cooldowns, optional daily-loss and drawdown halts per strategy; alerts on any wallet you add | Senpi |
| Following traders and strategies | A strategy marketplace: join a public group and your bot copies its owner’s signals, sized by your own limits; the owner earns the trading fees | Mirror a trader you choose, after the agent researches their record on whole-book profit, margin, drawdown, timing and fee load, sized 0.1x to 4x, with an exit attached | Tie |
| Custody | Non-custodial: you generate a Hyperliquid API key that can trade but not withdraw and add it to a bot; in the beta, credentials stay on your own dedicated server | Non-custodial embedded wallets with exportable keys; each strategy trades from its own funded strategy wallet, the agent holds no keys, and withdrawals are made by you in the app, behind MFA | Tie |
| Developer surface | The Signal API (19 endpoints), an Agent API, an MCP server, TradingView and n8n integrations, and Rust strategies in the beta | Senpi Connectors over MCP and the open-source Senpi Skills on GitHub for agents and scripts | Tie |
| Backtesting and paper trading | In the beta app: backtest on historical candles, then run the bot on live data with simulated capital before going live | No backtest or paper mode yet; the agent shows you the rules before anything opens and you can start live from $10 | Katoshi |
| Execution control, markets and accounts | Scale, grid and multi-take-profit orders, bulk actions, a Rust engine on a dedicated Tokyo server; perps, spot, HIP-3 markets, vaults and sub-accounts | Hyperliquid crypto perps plus the XYZ DEX stocks, pre-IPO names, metals, indices and FX, traded by the agent; no spot, no vaults | Katoshi |
| What it costs | No subscription. Current app: a builder fee from 0.060% of each perp trade on the first tier to 0.010% above $25M of 30-day volume; spot sells 0.6% to 0.1%; vault bots 1 credit a day. Beta app: prepaid server time in USDC plus 1 bp, or 0 bp with Katoshi’s referral code | Pro $50, Advanced $100, Quant $200 a month with AI credits; 0% Senpi trading fee on every paid plan, 0.03% on the free trial or without a paid plan; Hyperliquid’s own fees apply on top: 0.015% maker and 0.045% taker at the base tier on crypto perps, less at higher volume, and about 0.009% taker on most XYZ markets (GOLD, MSTR and three others cost 0.09%) (senpi.ai/fees) | - |
Winner column: Senpi, Tie, or - for rows that describe what each product is rather than which is better.
At a glance
Katoshi at a glance
An automation engine for Hyperliquid only. Bots connect a Hyperliquid account through an API key; TradingView alerts, webhooks and rules in Katoshi’s expression language fire orders through the Signal API, with scale, grid and multi-take-profit orders. Katoshi Agents add a language-model agent on your own OpenRouter key; an MCP server exposes the same trading tools to Claude Desktop or n8n. Public groups form a strategy marketplace whose owners earn the trading fees. A new beta app, which Katoshi says will replace the current one, runs Rust strategies with backtests, paper trading, logs and alerts on a dedicated Tokyo server. No subscription: a builder fee by volume tier, or prepaid server time plus 0 to 1 bp in the beta. Web only.
Senpi at a glance
An AI trading agent for Hyperliquid. Describe a strategy in plain language and the agent researches the market, writes it and runs it 24/7 on a hosted runtime under a deterministic risk harness; or mirror a trader you choose, after it researches their record. Quant Desk scores any Hyperliquid address on six dimensions; Senpi Signals track proven wallets against the crowd. Every strategy built from a rule or a template exits through DSL, a stop that ratchets up as the trade gains, and can carry daily-loss and drawdown halts on that strategy. Runs on web, iOS, Android, a Chrome extension on 28 sites, and inside Claude, ChatGPT, Grok, Hermes and Muse. Plans Pro $50, Advanced $100, Quant $200 a month, 0% Senpi trading fee on paid plans. Fund with a card, Apple Pay or Google Pay through the on-ramp, or with USDC on seven chains. Non-custodial wallets with exportable keys.
Quick verdict
| If you want | Choose |
|---|---|
| To run your trading agent from Claude, ChatGPT or your phone | Senpi |
| A perp strategy written from a sentence and run while you sleep | Senpi |
| Proven-wallet signals and a wallet score on 28 sites | Senpi |
| DSL on every strategy built from a rule or a template and optional halts per strategy | Senpi |
| To code a strategy, backtest it and paper trade it first | Katoshi |
| TradingView alerts executed on Hyperliquid, including spot and vaults | Katoshi |
What Katoshi is built for
Backtest and paper trade before going live
The beta app takes a Rust strategy through a backtest on historical candles and paper trading on live data with simulated capital, then to live execution, with logs for every decision.
TradingView and code straight into Hyperliquid
TradingView alerts, webhooks, n8n and a 19-endpoint Signal API turn any external signal into Hyperliquid orders, including scale, grid and multi-take-profit orders, across perps, spot, vaults and sub-accounts.
Low cost at volume, and nearly free in the beta
No subscription; the builder fee falls to 0.010% above $25M a month, and the beta charges 0 to 1 bp plus server time.
The agent, and where you reach it
Both products have an AI agent. The difference is who runs it and where it lives.
| Criterion | Katoshi | Senpi |
|---|---|---|
| The agent | Katoshi Agents: pick a model, write a system prompt, whitelist the trading actions; called by chat, API or webhook | Your Senpi agent researches, writes and runs strategies under a risk harness; it shows you the rules before anything opens |
| Model cost | Your own OpenRouter API key, billed by OpenRouter | Included: monthly AI credits on every plan |
| Inside AI assistants | An MCP server for Claude Desktop, n8n and other MCP clients, with the Signal API as tools | Senpi Connectors for Claude, ChatGPT, Grok, Hermes and Muse: one address and a Senpi sign-in |
| On the sites you read | None | Senpi Companion on 28 sites: Quant Desk on any wallet, the Signals brief, Trade the Take |
| Phone | Web only; push and email alerts | iOS and Android apps |
Senpi's edge here
Katoshi gives a builder a capable kit: an agent on the model of your choice and an MCP server with real trading tools. Senpi gives a trader a finished agent with its own runtime and harness, already inside the assistants and sites they use, with nothing to wire and no model bill on the side.
Senpi Signals vs Katoshi’s indicators
What each product tells you about the market before you trade.
| Criterion | Katoshi | Senpi |
|---|---|---|
| The read | 17+ indicators in the trigger language and chart analysis by the agent: RSI, MACD, Bollinger bands, EMA, SMA, ATR, OBV, VWAP | Senpi Signals: per asset, the biggest recent moves from proven wallets set against the broader cohort, scored, with the driver named |
| One wallet in depth | Your own bots’ equity, PnL, drawdown and win rate | Quant Desk: a score over six explained dimensions, a protection audit of the live book, leaks priced in dollars, on any Hyperliquid address |
| Alerts | A log of every signal your bots receive, with push and email notifications | An alert when any wallet you add opens, adds or cuts, plus the agent’s own read before it trades |
Senpi's edge here
Katoshi reads price; that is what a rules engine needs and it does it well. Senpi also reads the people behind the price: what wallets with a record are doing, against the crowd, and how good any address really is.
From an idea to a protected position
Building, testing, running and protecting a strategy.
| Criterion | Katoshi | Senpi |
|---|---|---|
| Starting point | A rule in code, ta.crossover(ta.ema($$, 9), ta.ema($$, 21)), a TradingView alert, or a Rust strategy | A sentence: “short ETH when funding flips positive and open interest is at a 7-day high, max 2% of equity” |
| Testing first | Backtest on historical candles and paper trade on live data in the beta | None yet; review the rules the agent shows you and start from $10 |
| Running | Bots execute signals 24/7; in the beta, on your own Tokyo server with logs and server monitoring | A hosted runtime runs it 24/7 under a risk harness; edits apply live |
| Exits and halts | SL and TP per signal, several partial TPs, max order size, coin whitelist, bot on/off | DSL on every strategy built from a rule or a template; cooldowns, optional daily-loss and drawdown halts per strategy; watch-only alerts on wallets you add |
Evenly matched
This one splits by who you are. If you write code, Katoshi’s backtest-then-paper-then-live path is the more rigorous one and Senpi has nothing like it yet. If you don’t, Senpi takes you from a sentence to a running, protected strategy, and its stops and optional per-strategy halts go further than per-signal SL and TP. Neither product can promise a fill at a stop’s trigger price.
Markets, accounts and cost
Where each one trades and what it charges.
| Criterion | Katoshi | Senpi |
|---|---|---|
| Markets | Hyperliquid perps, spot and HIP-3 markets (stocks, FX and more) | Hyperliquid perps plus the XYZ DEX stocks, pre-IPO names, metals, indices and FX; no spot |
| Accounts | Several accounts, sub-accounts and vaults from one dashboard | A main wallet and separate strategy wallets per strategy |
| Price model | No subscription: a builder fee by 30-day volume, 0.060% to 0.010%; in the beta, server time plus 0 to 1 bp | A flat plan: Pro $50, Advanced $100, Quant $200 a month, 0% Senpi trading fee |
| Cost of $150K a month of perps | About $90 in builder fees at 0.060% on the current app; in the beta, server time plus $0 to $15 | $50 on Pro, whatever the volume |
Winner: Katoshi
Katoshi trades more of Hyperliquid (spot and vaults too) and, in its beta, costs less than any subscription for most traders. On the current app, Senpi’s flat $50 is cheaper above roughly $85K to $100K of monthly volume. Hyperliquid’s own fees apply on both.
Where Senpi is stronger
An agent inside Claude, ChatGPT and your phone
Senpi Connectors put your own agent in Claude, ChatGPT, Grok, Grok Bot, Hermes and Muse, and the iOS and Android apps put it in your pocket, with no model key to bring.
Signals from proven wallets
Senpi Signals and Quant Desk read what wallets with a record are doing against the crowd and score any Hyperliquid address, on senpi.ai and on 28 sites through the Companion.
Strategies written, run and protected for you
A sentence becomes a running strategy under a risk harness, with DSL on every strategy built from a rule or a template and optional daily-loss and drawdown halts per strategy.
Which should you choose?
Choose Katoshi if
- You write your own rules or Rust and want to backtest and paper trade before risking money
- You trade from TradingView alerts or your own scripts and need them executed on Hyperliquid
- You trade spot or run Hyperliquid vaults and sub-accounts
Choose Senpi if
- You want your trading agent inside Claude, ChatGPT or your phone
- You want a Hyperliquid perp strategy written from a sentence, run and protected for you
- You want proven-wallet signals and a wallet score on the Hyperliquid wallets you follow
Conclusion
Senpi is the agent. It writes the strategy from a sentence, runs it 24/7, protects every strategy built from a rule or a template with DSL and optional per-strategy halts, and comes to you inside Claude, ChatGPT, Grok, Hermes or Muse, on your phone, or on the Companion while you read X or TradingView. Katoshi’s agents are a framework you assemble; Senpi’s is one you talk to.
Senpi Signals and Quant Desk add what a rules engine does not have: a scored read of what proven wallets are doing against the crowd, and a six-dimension score on any address, with the leaks in your own book priced in dollars.
Senpi’s price is flat, $50 a month on Pro with 0% Senpi fee, and above roughly $85K to $100K of monthly perp volume it costs less than the builder fee on Katoshi’s current app.
Katoshi is the better tool for a trader who codes: TradingView and webhooks straight into Hyperliquid, spot and vaults, and in its beta a backtest-to-paper-to-live path at almost no fee. If you write the logic, Katoshi; if you want the logic written, run and protected for you, Senpi.
Frequently asked questions
What is Katoshi?
An automation engine for Hyperliquid. It executes trading signals from TradingView alerts, webhooks, its own trigger language and AI agents through a Signal API, and runs a strategy marketplace. A new beta app runs Rust strategies with backtests and paper trading on a dedicated server.
Is Katoshi safe?
Katoshi is non-custodial: you connect a Hyperliquid API key, which can place trades but cannot withdraw, and approve a builder fee. In the beta app, credentials and strategies stay on your own dedicated server. As with any automation, a bad rule or signal can still lose money.
What are Katoshi’s fees?
On the current app, a builder fee by 30-day volume, from 0.060% of each perp trade on the first tier to 0.010% above $25M, plus 0.6% to 0.1% on spot sells; vault bots cost 1 credit a day. The beta app charges prepaid server time plus 1 bp, or 0 bp with Katoshi’s Hyperliquid referral code. Hyperliquid’s own fees come on top.
Does Katoshi work with TradingView?
Yes. A TradingView alert sends a webhook to Katoshi’s Signal API, which places the order on Hyperliquid. Senpi does not execute TradingView alerts; its Companion brings Signals and Quant Desk onto TradingView pages instead.
Can Katoshi’s AI agents trade for me?
Yes, within the actions you whitelist. A Katoshi Agent runs on a model you choose through your own OpenRouter key and can be called by chat, API or webhook. Senpi’s agent needs no model key and runs strategies 24/7 under a risk harness.
Katoshi or Senpi: which is the better AI trading agent for Hyperliquid?
For a trader who wants the strategy written, run and protected from a sentence, inside Claude, ChatGPT or a phone, Senpi. For a developer who wants to code the logic, backtest it and wire TradingView into Hyperliquid, Katoshi.
Does Katoshi have backtesting and paper trading?
In its beta app, yes: backtests on historical candles and paper trading on live data before going live. Senpi has no backtest or paper mode yet.
Who should not pick Senpi?
Anyone who wants to backtest or paper trade first, trade spot or run vaults, or execute TradingView alerts: Katoshi does those and Senpi does not. And anyone trading under roughly $85K a month, where Katoshi’s builder fee costs less than a $50 Pro plan.
Can I run Senpi from Claude or ChatGPT?
Yes. Senpi Connectors put your Senpi agent inside Claude, ChatGPT, Grok, Hermes and Muse: you ask it to research a market, build or edit a strategy, review a wallet or close a position, and it does it on Hyperliquid without you opening the app.
What is DSL?
Senpi’s dynamic stop loss: a stop attached to each position in a strategy Senpi runs, which ratchets up as the trade gains and never loosens, so a winner gives back only part of its gain before it closes. No stop can guarantee a fill at its trigger price.
Related comparisons
- Keel vs SenpiAI agents and strategy builders for Hyperliquid
- WunderTrading vs SenpiRule-based bot platform
- fomo vs SenpiSocial trading app