# DXAP Docs > DXAP puts an AI agent on your Hyperliquid account. You write the strategy and set the limits, it trades around the clock, and you can read every decision it makes. DXAP is built by DX Research Group (DXRG). --- # DXAP documentation Source: https://docs.dxap.ai/ DXAP / Hyperliquid / Alpha From strategyto agent. Describe your strategy. Define its trading limits. Follow your agent's decisions, positions, and trades in one workspace. Get started → Open DXAP ↗ Put your strategy to work with an autonomous AI trading agent on Hyperliquid. Choose its model, define its approach, and set its trading boundaries. Your agent reviews market and account context on its schedule and triggers, deciding when to act and when to wait. Follow its decisions in the workspace and refine its direction through chat. ## Learn the workflow ## How DXAP fits together | You configure | DXAP uses it to | | --- | --- | | **Strategy** | Give the agent its objectives and decision criteria. | | **Account** | Identify the Hyperliquid account whose positions and funds it trades. | | **Model and schedule** | Choose the reasoning model and the regular evaluation interval. | | **Trading policies** | Check proposed orders against market and notional limits. | | **Instructions** | Carry confirmed chat direction into future agent turns. | DXAP currently offers live trading in alpha. A referral code is required for new-user access. Start with [account setup](/guides/connect-account/) and [fees and risks](/reference/fees-and-risks/) before authorizing an agent. Funds stay in your Hyperliquid account; an agent wallet grants trading authority. That authority can still expose your funds to trading losses. --- # Getting Started Source: https://docs.dxap.ai/getting-started/ Your first goal is to set up an agent you understand and review what it does. You do not need an API key or a local development environment to use DXAP. ## Before you begin - Have a supported wallet available for sign-in and account approvals. - Have a referral code for new-user alpha access. Visit [DXAP](https://dxap.ai) to request access if needed. - Review [agent wallets](/reference/agent-wallets/) and [fees and risks](/reference/fees-and-risks/). :::caution[Live trading] Your agent trades through your connected Hyperliquid account. Review its strategy and trading limits before launch. Trading can lose money; the account minimum is not a suggested trading budget. ::: ## 1. Prepare an account Sign in and open **New agent**. Select the Hyperliquid account you want the agent to trade. Account setup walks through **Terms**, **Fund**, **Allow trading**, and **Allow fees**. The current setup requires at least **10 USDC** in the selected account before an agent can run. See [Connect an account](/guides/connect-account/) for what each approval means. ## 2. Build and review Choose **Guided** to answer questions and choose a starting point, or **Custom** to describe your own strategy. Review the generated card: strategy, model, schedule, markets, and trading limits. **Save card changes** saves the draft. **Start the agent** launches it. Read [Create an agent](/guides/create-an-agent/) before starting. ## 3. Read the first turn Open the agent in the workspace. Review **Activity**, then compare it with **Positions** and **Trades**. A completed turn may decide not to trade. A proposed or submitted order is not proof of a fill. Use [Review activity](/guides/review-activity/) to understand the record. ## 4. Know how to stop You can pause an agent in its settings or revoke its trading key on Hyperliquid. Closing the browser does not stop it. Pausing and revoking access do not close existing positions or cancel resting orders. See [Pause and revoke access](/guides/pause-and-revoke/). --- # Connect an account Source: https://docs.dxap.ai/guides/connect-account/ Sign in to [DXAP](https://dxap.ai) with your wallet, open **New agent**, and select the Hyperliquid account the agent will use. Check the selected account address before continuing. ## Complete account setup | Step | What to review | | --- | --- | | **Terms** | Read and accept the terms presented in the application. | | **Fund** | Follow the funding flow for the selected account. Setup currently requires at least 10 USDC before an agent can run. | | **Allow trading** | Authorize the agent wallet that DXAP uses to place and cancel orders. | | **Allow fees** | Review and approve DXAP's builder fee on Hyperliquid. | Sign-in, trading approval, and fee approval serve different purposes. Being signed in does not mean the account is ready to trade. Confirm the setup checks are complete before moving to the strategy card. ## Understand the account boundary The account holds the balance and positions. The agent wallet is the signing key with trading authority. For balances and positions, inspect the selected account address, not the agent wallet address. See [Agent wallets](/reference/agent-wallets/). If the application offers subaccounts, availability depends on the connected account's eligibility. Follow the choices presented in account setup. ## If setup is incomplete - Confirm the connected wallet and selected account are the ones you intended to use. - Check which setup step is still pending; review the status after completing the wallet action. - If funding is pending, confirm the balance in Hyperliquid before attempting to launch. - If trading or fee approval is missing, complete the matching approval in the app. Next: [Create an agent](/guides/create-an-agent/). --- # Create an agent Source: https://docs.dxap.ai/guides/create-an-agent/ Turn your trading idea into an AI agent. Complete [account setup](/guides/connect-account/), then use the builder to shape your strategy and choose how your agent will operate. ## Choose a build method | Method | Flow | | --- | --- | | **Guided** | Answer four short questions, choose a starting point, then refine the proposal in chat. | | **Custom** | Describe your strategy in your own words or edit its configuration directly. | Both paths end at **Review and launch**. A starting point is an editable proposal, not evidence that the strategy will perform well. ## Make the strategy specific Explain which markets to consider, the conditions for taking action, and when the agent should wait. Keep the written strategy consistent with the market and trading policies on the card. Writing a limit in the strategy is different from configuring a policy field. Use the actual [agent settings](/reference/configuration/) to set enforced limits. ## Review the card Check the selected account, agent name, model, strategy, evaluation schedule, market access, and entry and position limits. Refine the proposal in chat or edit the fields, then review the resulting card again. Example account from the DXAP walkthrough. The values illustrate the interface; they are not recommended settings or a performance record. ## Save or launch Use **Save card changes** to keep refining your draft at your own pace. When you've reviewed the strategy and its limits, choose **Start the agent** to launch it. After launch, check your agent's settings and next scheduled turn in the workspace. Your agent runs autonomously from there, following its schedule and triggers even when your browser is closed. [Follow its activity](/guides/review-activity/) to see what it decides and how its trades unfold. --- # Review activity Source: https://docs.dxap.ai/guides/review-activity/ Follow your agent's decisions as they unfold. Select an agent and check its trading account, then open **Activity**. Each turn is a fresh review of its strategy against the available market and account context. ## Read the record in order 1. Open **Activity** and select the turn you want to inspect. 2. Read its summary and the available decision and action details. 3. Compare the outcome with **Positions** and **Trades** for that account. 4. If you need more context, ask the selected agent in chat about the decision. ## Distinguish decisions from fills | What you see | What it establishes | | --- | --- | | A completed turn | The evaluation finished. It may have decided to wait. | | A proposed action | The agent attempted an action; policy and execution checks still matter. | | A rejection or failure | That action did not complete as requested. Inspect the reason. | | A submitted order | An order was sent; inspect its status rather than assuming it filled. | | A fill in Trades | An execution was recorded. Compare its size and price with the decision. | A no-trade decision can be consistent with a strategy whose conditions were not met. Repeated failures need a different response: check account readiness, policy limits, and the actual error details. ## Ask a focused question Try asking, “What conditions were missing in the last turn?” or “Explain the action in this Activity entry.” Chat can help interpret the available record; verify execution outcomes against the trading views. For updates in Telegram, connect your account and use `/last`, `/positions`, and `/triggers`. See [Use Telegram](/guides/telegram/). To change future behavior, use [Chat and instructions](/guides/chat-and-instructions/). --- # Chat and instructions Source: https://docs.dxap.ai/guides/chat-and-instructions/ Talk directly with your selected agent. Ask why it made a decision, explore its strategy, or work through a change together. The conversation draws on that agent's strategy, account, and recent activity. ## Ask before changing Start with “Walk me through your last decision” or “What would make you consider your next entry?” Use the conversation to understand how your agent approaches its strategy. Chat itself doesn't place a trade, trigger a trading turn, or approve a settings change. ## Review a settings proposal Refine your strategy together. Your agent can propose changes to its settings or approach, and you decide which to approve. Review the proposal, confirm the changes you want, then check the updated **Agent settings**. Builder chat works on the draft strategy card. Changing that draft still requires the separate launch action described in [Create an agent](/guides/create-an-agent/). ## Confirm a persistent instruction Give your agent direction it can carry forward. It can turn your request into a proposed instruction for future turns. Review and confirm it to make it active, and cancel it whenever it no longer fits your strategy. An instruction is strategy context. It does not replace the configured trading policies or the account's trading authorization. ## If the conversation is interrupted Reopen the conversation and inspect the saved messages and proposal state before repeating an approval. Confirm whether the change already appears in the agent's saved settings. --- # Pause and revoke access Source: https://docs.dxap.ai/guides/pause-and-revoke/ Your agent runs autonomously, following its schedule and triggers while you're away. When you want it to stop, you can pause it in DXAP or revoke its trading authority on Hyperliquid. ## Pause in DXAP Open **Agent settings**, choose **Edit**, change **Status** from **Active** to **Paused**, and save. Check that your agent now shows **Paused**, then review recent activity for anything already in progress. Before resuming, review the account, positions, resting orders, strategy, schedule, and trading policies. ## Revoke on Hyperliquid To remove your agent's permission to trade, open your account's [Hyperliquid API settings](https://app.hyperliquid.xyz/API) and revoke the relevant agent wallet. Revoking the key removes its trading authority; pausing in DXAP leaves that permission in place. ## Review open exposure separately Neither pausing nor revoking is a close-all action. Existing positions and resting orders remain on the venue. Inspect them in Hyperliquid and manage them separately. | Action | Effect | | --- | --- | | Close the browser | Your agent continues running autonomously. | | Pause the agent | Pauses your agent while keeping its trading permission in place. | | Revoke the key | Removes the key's permission to trade on your account. | | Close a position or cancel an order | A separate trading action on the venue. | See [Agent wallets](/reference/agent-wallets/) for the distinction between the account and the signing key. --- # Use Telegram Source: https://docs.dxap.ai/guides/telegram/ Stay in touch with your agent through Telegram. Receive updates, check positions and recent decisions, change notification preferences, or request a turn with an instruction that applies to that turn only. ## Connect your Telegram account 1. Sign in to [DXAP](https://dxap.ai) and open **Profile > Settings**. 2. Under **Notifications**, choose **Connect Telegram**. 3. Choose **Open Telegram**, or scan the QR code with your phone. Use the link generated for your account. 4. In the bot's private chat, tap **Start** and wait for the account-linked confirmation. 5. Return to Settings and choose **I've linked Telegram** or **Refresh** if the connection has not appeared yet. If the link expires, choose **Generate new code** in Settings and open the new link. Commands work in your private chat with the bot. :::note[Connected does not mean alerts are on] Linking Telegram enables commands, including notification controls and confirmed turn requests. Choose which alerts to receive for each agent using `/notifications`. ::: ## Choose your alerts Send `/notifications`, then tap the option you want: | Option | What you receive | | --- | --- | | **Agent updates** | Summaries of your agent's finished turns. | | **Trade alerts** | Notifications of recorded agent trades. | | **Both** | Agent updates and trade alerts. | | **Off** | No automatic alerts for this agent. You can still use commands. | The selection updates that agent's notification preferences. Send `/status` to check the current settings and the last alert sent. With several agents, configure each one separately, for example `/notifications Owl`. ## Slash commands With one agent, send the command on its own. With several agents, add the agent's name where shown. Square brackets below indicate optional text; do not type the brackets. | Command | What it does | | --- | --- | | `/positions [agent name]` | Shows open positions, direction, leverage, size, entry price, and unrealized PnL in USD. | | `/last [agent name]` | Shows the latest finished turn, including failed or cancelled turns, and whether another turn is pending or running. | | `/triggers [agent name]` | Shows the schedule and next scheduled time separately from active trigger conditions. | | `/status [agent name]` | Shows the Telegram connection, agent alert settings, and last alert sent. Without a name, lists your agents' status. | | `/agents` | Lists your agents, trading mode, activity state, and alert status. | | `/notifications [agent name]` | Opens the buttons for choosing agent updates, trade alerts, both, or off. | | `/run [instruction]` | Previews a manual turn for your single agent, with an optional instruction. Requires confirmation. See the examples below for several agents. | | `/help` | Shows the command list and usage examples. | | `/start` | Shows setup guidance. Opening the connection link from Settings uses this command to link your account. | Use the full agent name shown by `/agents`. Names are matched without regard to capitalization. If two agents have the same name, rename one in the app before selecting it through Telegram. When a reply offers another page, follow its command, such as `/positions Owl page=2` or `/triggers Owl page=2`. Positions are shown five per page. Times in replies are in UTC. ## Read positions and triggers `/positions` uses the latest recorded position data and prices. Check the timestamps: unrealized PnL can change before the next update. It excludes realized PnL, fees, and funding. Use **Positions** and **Trades** in the app to review the account and execution outcomes. `/triggers` separates **Schedule** from **Triggers**, so an hourly schedule and a price condition are easy to distinguish. Triggers only fire when your agent is not paused. Use the app to change schedules or trigger conditions. ## Request a turn For a single agent, send `/run` to request a turn using its existing strategy. To add an instruction for that turn, include it after the command: ```text /run Review my existing positions before considering new entries. ``` With several agents, use the agent's name. Separate the name from an instruction with `|`: ```text /run Owl /run Owl | Review my existing positions before considering new entries. ``` Review the agent name and **Directive** in the preview. Tap **Confirm** within five minutes to queue the turn, or **Cancel** to dismiss it. A queued turn may result in trades; the confirmation acknowledges the request, not a completed trade. The directive applies to that turn only. If you leave details unspecified, the agent interprets the instruction under its existing settings and trading limits. It does not update the agent's saved strategy or create a persistent instruction. For those changes, use [Chat and instructions](/guides/chat-and-instructions/) in the app. A paused agent must be resumed in the app before you can request a turn. Manual turns share a 30-minute cooldown with the app, and another manual turn cannot be requested while one is pending or running. The bot explains when a request is unavailable. Use `/last` and the app's **Activity** to review the result. ## If something is missing - **Connected, but no alerts:** Send `/status`, then use `/notifications` to enable the categories you want for the correct agent. - **A button has expired or settings changed:** Send `/notifications` or `/run` again and use the latest reply. - **The bot asks you to link again:** Return to **Profile > Settings**, generate a new connection link, and complete the linking flow. - **Positions look old or have no recorded data:** Check the timestamps and review the account in the app. Missing data does not establish that the account has no positions. To disconnect, open **Profile > Settings** and choose **Disconnect** in the Telegram section. This removes Telegram access and alerts; your agent continues running. To stop future agent activity, see [Pause and revoke access](/guides/pause-and-revoke/). --- # Season 0 Chapter 0: Exploration Source: https://docs.dxap.ai/guides/season-0-exploration/ **Season 0 · Chapter 0: Exploration** is our first themed way to watch your agents. Agent Catacombs visualizes your agent’s trades, research and quiet stretches as a dungeon story. Open the **???** tab in DXAP, or use the Season 0 link on Profile. Choose your agent, watch its journey, and share a 30-second recap on X. ## All core functionality is unchanged **This is a visualizer, not a separate game.** Watching does not place orders, change policies, alter schedules, steer agents or generate real returns. Seeded visitors, cat memes and discoveries are fictional. Confirmed trading activity and server-awarded achievements supply the real milestones. Future chapters can introduce different ways to watch and share your agents. ## Watch the journey Choose your agent to see its real activity interpreted as a dungeon crawl. Entries become encounters; closing trades become victories, escapes and treasure scenes. Research and other recorded actions have their own effects. Familiar crypto characters and cat memes appear as parody visitors. A live X dispatch shows its source post separately from fictional dialogue. The journey keeps moving between turns, even when an agent waits hours before acting. Routes and discoveries are seeded from the agent’s history and time, so returning to the same state produces a consistent scene. These encounters never invent trades or increase your volume, results or achievements. As your level rises, the environments become stranger: glass gardens, the DXRG Mirror Observatory, cat reefs, an upside-down sea, recursive archives, silicon dreams and the Unreasonable Exchange. New destinations continue unlocking beyond level 100. DXAP and DXRG standards mark the passages. Sound is optional, with separate music and effects controls. The viewer works on desktop and mobile, respects reduced-motion preferences, and pauses background rendering when its tab is hidden. Closing the viewer never stops your agent from trading. ## Share a 30-second recap From Profile, choose **Share as dungeon crawler**, or use **Share this week** in the viewer. Choose the last 24 hours or this week, then create the video. Desktop defaults to landscape for X; portrait is also available. You can include an access code. The video and prepared post both display **dxap.ai**. Supported browsers can create the full 30-second recap faster than playback. Other browsers record it in real time; keep the tab visible when prompted. The finished video always plays at normal speed with the selected sound settings. **Save video & share on X** downloads the video and opens a prepared draft. Attach the saved video in X before posting. Nothing is posted automatically. **Copy replay link** shares the captured summary and selected highlights, including the wallet achievements available at capture. A partial history is labeled. The underlying live agent and core trading controls are unchanged. We are still experimenting with ways to turn watching your agents into entertainment. Tell us what is useful, what gets repetitive and what you want to see next. ## Dates and levels Trading achievements begin **September 22, 2026 at 00:00 UTC**. Chapter 0 lasts two weeks, from September 22 through October 6. Its countdown starts at **14 days left in Chapter 0**. The two-week bonus window ends **October 6, 2026 at 00:00 UTC**. The level follows your wallet’s eligible Hyperliquid **mainnet volume carrying the DXAP builder code across your linked accounts and agents**. The builder code is the fee attribution on an order routed through DXAP; it is separate from the referral codes you share with friends. Your agent’s fills count. If you manually close or partly close a position through DXAP, that fill counts when it carries the DXAP builder code too. A trade made elsewhere on Hyperliquid without the DXAP builder code does not count. Each filled entry and exit counts once; forced-liquidation volume does not count. Level 100 corresponds to $1M; levels continue higher after that milestone. An incomplete volume history is displayed as a lower bound. Recorded examples are labeled and do not earn live achievements. ## Achievements DXAP returns the achievements you currently hold. The six current profile badges are First agent, Squad of five, On the board, Top 20, Live account and Referrer. They reflect current account state and can disappear when their conditions no longer hold. The permanent trading milestones cover DXAP builder-code volume of $1K, $10K, $100K, $500K and $1M; closing-fill returns of +1%, +10% and +100%; and closing-fill losses of 1%, 10% and 50%. Partial closes qualify. There are no separate streak, leverage or liquidation achievements. The same backend closing-fill percentage is used for the trade display, P&L cards and profit/loss awards; unavailable history stays unavailable. Profile, Summary and the visualizer show backend awards. A saved recap includes its captured wallet achievement list, clearly separated from the selected agent’s trading results. New awards can arrive after the API’s short refresh delay. ## Team-distributed end-of-period bonuses :::note[Distributed by the DXAP team] Your DXAP team reviews eligibility and distributes the bonuses and referral refresh. The visualizer does not credit points or issue codes. ::: The DXAP team awards achievement bonuses **after the two-week period ends**. There is **no claim step**. Each qualifying permanent achievement stacks once per wallet. Bonuses use backend-confirmed achievements from eligible activity in the two-week window. Existing profile badges are displayed but are not used as undated seasonal bonus evidence. The team checks late-reconciled activity during the final award audit. | Achievement | Threshold | Bonus points | Single-use codes | | --- | --- | ---: | ---: | | First thousand | $1,000 DXAP builder-code volume | 1,000 | 1 | | Five figures | $10,000 DXAP builder-code volume | 10,000 | 1 | | Six figures | $100,000 DXAP builder-code volume | 150,000 | 2 | | Half a million | $500,000 DXAP builder-code volume | 750,000 | 3 | | Million traded | $1,000,000 DXAP builder-code volume | 2,000,000 | 5 | | In the green | Closing-fill return ≥ +1% | 250 | 0 | | Double digits | Closing-fill return ≥ +10% | 500 | 0 | | Doubled | Closing-fill return ≥ +100% | 1,000 | 1 | | First bruise | Closing-fill loss ≥ 1% | 500 | 0 | | Tough close | Closing-fill loss ≥ 10% | 1,000 | 1 | | Half down | Closing-fill loss ≥ 50% | 2,000 | 2 | The volume amounts are **extra points for each tier reached**, not a total balance or a replacement for normal volume points. They stack once per wallet across the chapter. For example, reaching $1M of eligible chapter volume qualifies for all five volume tiers, totaling **2,911,000 direct bonus points**. No one needs to force a trade to keep an already-earned tier. Loss milestones deliberately receive **twice the points** of the corresponding profit tiers, with more referral codes at the deeper thresholds. Bonuses are fixed, one-time recognition amounts; they do not increase with dollars lost or repeat on every losing trade. The maximum direct achievement bonus is **2,916,250 points and 16 owned, single-use referral codes per wallet**, if every listed achievement qualifies. Normal referral propagation remains unchanged, so referral-derived points are separate from this direct-award cap. Points are program points, not cash or a refund of trading losses. ## Invite access and second-week refill At the weekly allocation review, signed-up accounts with **zero usable referral invites** receive two to start. This is a reviewed grant, **not an automatic grant at signup**. Prior claims remain counted. During the second-week review, an account whose **current allocation has been redeemed at least twice by new users** has its usable invite allowance restored to that allocation’s full recorded size. A shared but unused code does not count. The team verifies the claims and existing allowance before making the one-time refill. There is **no points-balance requirement** for this refill. It is separate from the achievement bonuses awarded after the two-week period. ## Why these bonuses? Early volume milestones make the experiment accessible. Larger milestones recognize sustained use. Loss awards are intentionally more generous than winning awards. Fixed amounts and one-time eligibility keep the same badge from producing repeated bonuses. The invite refill supports people who have brought at least two others into the community. All standard [fees and risks](/reference/fees-and-risks/) and existing referral terms remain unchanged. The visualizer never alters account permissions or trading behavior. ## Tell us what you think Share your replay, tell us what feels repetitive, and suggest future themes. [Open DXAP](https://app.dxap.ai/lab) or [give feedback in Discord](https://discord.gg/huJ5fZGvRD). --- # Agent settings Source: https://docs.dxap.ai/reference/configuration/ Review the card before launch and the saved settings whenever you change an active agent. ## Identity and evaluation | Setting | Meaning | | --- | --- | | **Name** | How you identify the agent in the workspace. | | **Account** | The Hyperliquid execution account the agent trades. | | **Model** | The reasoning model selected from the app's available models. | | **Strategy** | The agent's written objectives and decision criteria. | | **Schedule** | The regular evaluation interval. The current builder accepts 30 minutes to 4 hours. | | **Schedule instructions** | Additional context for scheduled evaluations. | | **Status** | Whether the saved agent is Active or Paused. | A schedule is an evaluation interval, not a promise to place an order at each interval. Triggers can also initiate agent evaluations. ## Market access Market scope controls which parts of Hyperliquid's market universe the agent may use: core markets, HIP-3 markets, or a selected HIP-3 DEX where available. Within that scope, the symbol policy can allow all supported symbols, allow only listed symbols, or block listed symbols. The editor accepts one of these modes at a time. Use the market selector's supported symbols rather than guessing names. ## Trading limits In **Agent settings**, choose **Edit** and find **Risk Limits**. Save your changes to apply them. | UI field | Controls | | --- | --- | | **Max Entry Notional** | The maximum notional value of an individual opening trade. | | **Max Position Notional per Symbol** | The projected position notional allowed for a symbol when opening exposure. | | **Max open positions** | The maximum number of distinct open symbols on the selected account, including manual positions and pending entries. | | **Re-entry cooldown (minutes)** | How long to wait after a full close before opening the same symbol again on the selected account. | | **Total Entries per 24h** | The opening-trade count allowed over a rolling 24-hour window. | | **Total Entry Notional per 24h** | The opening-trade notional allowed over a rolling 24-hour window. | Notional is the size of the exposure in USD terms. It is not the same as the margin committed or the maximum amount you can lose. Rolling limits are not calendar-day allowances. ### Max open positions Enter a positive whole number, or leave the field blank for no limit. The cap counts distinct symbols, so adding to an existing position does not use another slot. Reducing or closing a position remains allowed by this limit. Lowering the cap below the current count does not force positions to close. For example, with a cap of **1** and a BTC position already open, the agent cannot open ETH. It can still increase, reduce, or close BTC, subject to other policies. This is separate from the rolling 24-hour entry limit, which counts opening and increasing orders. ### Re-entry cooldown Enter a positive whole number of minutes, or leave the field blank to turn the rule off. For example, **60** means one hour. Any confirmed full close starts the timer, including a profitable exit, stop, manual close, or liquidation. During that period, the agent cannot open a new long or short position in the same symbol on that account. Other symbols are unaffected. Partial closes do not start the timer. The rule permits increasing, reducing, and closing already-open positions, subject to other policies. The timer starts from the closing fill's execution time; activity outside the app is recognized when its fills are received. Changing the duration applies to the last recorded full close immediately. Configured policy fields and strategy text have different roles. The strategy guides the model; the [policy checks](/reference/policy-and-execution/) constrain execution. Review the actual saved fields rather than relying on an instruction in chat. --- # Agent wallets Source: https://docs.dxap.ai/reference/agent-wallets/ DXAP uses Hyperliquid agent wallets for trading authorization. Your balance remains in your Hyperliquid account; DXAP does not hold that balance. ## Account versus key | Item | Role | | --- | --- | | Your connected wallet | Signs you in and approves account actions. | | Hyperliquid account | Holds the balance, positions, and orders being traded. | | Agent wallet / API wallet | A separate key approved to sign trading actions for the account. | | DXAP API key | An application credential for scoped DXAP API access; a different kind of key. | The agent wallet can place and cancel orders. It does not give DXAP withdrawal authority. Trading permission still carries risk: an authorized agent can place trades that lose funds. Hyperliquid also calls agent wallets **API wallets**. When checking balances or positions, use the actual account address. The signing key's address is not the account holding those funds. See [Hyperliquid's API wallet documentation](https://hyperliquid.gitbook.io/hyperliquid-docs/for-developers/api/nonces-and-api-wallets). ## Remove authority Revoke the approved key in [Hyperliquid API settings](https://app.hyperliquid.xyz/API). Pausing inside DXAP is a separate control. Neither action closes existing positions or cancels resting orders; see [Pause and revoke access](/guides/pause-and-revoke/). --- # Policy and execution Source: https://docs.dxap.ai/reference/policy-and-execution/ The model uses the strategy and available context to decide what to do. Execution applies configured checks before sending a proposed order to the venue. ## The decision path 1. A schedule or trigger makes an agent turn eligible to run. 2. The agent evaluates its strategy, account state, and available market context. 3. It can decide to wait or request an action. 4. Execution checks account authorization and the policies relevant to the action. 5. Submitted orders and execution outcomes are recorded for review. ## What the trading policy checks The current Hyperliquid path checks market scope and symbol permissions. For opening trades it also evaluates entry notional, projected position notional, and configured rolling entry limits. When configured, the open-position cap limits new symbols, and the re-entry cooldown blocks reopening a symbol after a full close. Requests that breach these checks are rejected. Opening limits should not be interpreted as identical rules for every action: closing exposure is distinct from opening it. See [Agent settings](/reference/configuration/) for the user-facing fields. ## What a successful check means A successful check means that request passed the applicable checks at execution time. It does not establish that the venue filled the order, that the fill matched the expected price, or that the trade will be profitable. Policies are not a guaranteed loss cap. Price changes, leverage, existing exposure, and venue execution still matter. Confirm what actually happened in [Activity and Trades](/guides/review-activity/). --- # Fees and risks Source: https://docs.dxap.ai/reference/fees-and-risks/ ## Fee categories DXAP's current published alpha pricing is **2.5 basis points (0.025%)** on filled volume routed through DXAP with its builder code: **$0.25 per $1,000 filled**. This includes your agent's trades and manual closes or partial closes made through DXAP. Filled openings and closings count when they carry the DXAP builder code; unrelated trades made elsewhere on Hyperliquid do not incur the DXAP builder fee. The current offer has no DXAP subscription or separate model charge. Review the live application's fee approval before authorizing it; see the [DXAP FAQ](https://dxap.ai/docs/faq). DXAP's Hyperliquid builder address is `0x5803667556F289aE854C914E59a2395B75C8aa24`. The fee approval you sign in the app names this address, and you can look up the fills and fees routed through it on Hyperliquid's public data. Hyperliquid trading fees and funding are separate. A position's trading result should be interpreted alongside its fees and funding, not just price movement. See [Hyperliquid's fees documentation](https://hyperliquid.gitbook.io/hyperliquid-docs/trading/fees). ## Account minimum The current onboarding flow requires at least **10 USDC** in the selected account before an agent can run. This is an account-readiness check, not a recommendation, an order-size guarantee, or a maximum-loss limit. ## Risks that remain - **Market and liquidation risk:** perpetual positions can lose money, including through liquidation. - **Model risk:** an agent can use incomplete context, misunderstand a strategy, or make a poor decision. - **Execution risk:** an order can fail, remain unfilled, or execute at a different price than expected. - **Operational risk:** alpha software and external services can be unavailable or change behavior. Trading policies constrain specific actions; they do not make trading safe or profitable. Non-custodial trading protects the custody boundary, not the account from trading losses. Past performance is not a guarantee of future outcomes. Know how to [review activity](/guides/review-activity/) and [pause or revoke access](/guides/pause-and-revoke/) before starting an agent. --- # FAQ Source: https://docs.dxap.ai/reference/faq/ ## What can I trade? Your agent can trade Hyperliquid perpetuals, including HIP-3 markets where available. Explore the market selector, then choose which markets your agent can trade through its market and symbol settings. ## Do I need an invitation? DXAP is in alpha and requires a referral code for new-user access. Visit [DXAP](https://dxap.ai) to request access. ## Can I paper trade or backtest first? At this time, DXAP supports live trading on Hyperliquid. You can build and refine your strategy, review its settings, and choose when to start your agent. ## Does saving a strategy start trading? No. Take your time building and refining your strategy. Saving the card keeps your draft ready; your agent launches when you choose **Start the agent**. ## Why did an active agent not trade? Waiting can be part of the strategy. Your agent may decide that the right conditions haven't arrived. An order can also be blocked by a trading limit or fail to execute. Open **Activity** to see what happened, then check **Positions** and **Trades** for the outcome. ## Will closing my browser stop the agent? No. Your agent runs autonomously, following its schedule and triggers even when you're away or your browser is closed. To stop your agent, you can pause it in **Agent settings** or revoke its trading authority on Hyperliquid. Existing positions and resting orders stay open until you manage them separately. See [Pause and revoke access](/guides/pause-and-revoke/). ## Does chatting with an agent place a trade? Use chat to explore your agent's thinking, refine its strategy, and give it direction. Chat itself doesn't place trades or trigger a trading turn. If your agent proposes a settings change or a persistent instruction, you review and approve it before it takes effect. ## Where do I find my agent's limits? Open **Agent settings** to see the trading boundaries you've set. The [settings reference](/reference/configuration/) explains each control, including how notional limits differ from margin or loss. --- # Glossary Source: https://docs.dxap.ai/reference/glossary/ | Term | Meaning in DXAP | | --- | --- | | **Agent** | An autonomous AI that reviews market and account context and makes trading decisions using your strategy and trading policies. | | **Agent wallet** | The Hyperliquid API wallet authorized to sign trading actions. | | **Builder** | The Guided or Custom workflow for drafting and launching an agent. | | **Builder fee** | The venue mechanism used to collect DXAP's trading fee; separate from the agent builder UI. | | **Execution account** | The account whose balance, positions, and orders the agent uses. | | **Fill** | An executed portion of an order. | | **HIP-3** | Hyperliquid's builder-deployed perpetual markets, shown by DEX where supported. | | **Instruction / directive** | Confirmed direction carried into future agent turns. | | **Notional** | The USD value of exposure; distinct from the margin committed. | | **Perpetual / perp** | A derivative position without a fixed expiry. | | **Policy** | A configured rule checked at execution time. | | **Schedule** | The agent's regular evaluation interval. | | **Strategy** | Written decision criteria and objectives given to the agent. | | **Trigger** | A configured condition that can initiate a turn. | | **Turn** | One agent evaluation, which may or may not result in trading. | --- # Releases Source: https://docs.dxap.ai/releases/ Product changes, improvements, and fixes, listed by version with the newest first. Entries marked **Upcoming** are awaiting deployment. ## 0.9.51 **October 3, 2026** More precise technical checks directly in agent chat, building on the expanded TA toolkit in 0.9.50. ### Added - **Direct candle statistics in agent chat.** Ask for completed-candle trend, breakout, range, or volume measurements directly in your selected agent's chat. Matching checks can cover up to **20 symbols per request**, preserving each symbol's values, window boundaries, and data coverage. ### Improved - **Focused technical lookups.** Chat uses candle statistics for exact window calculations and its existing indicator tools for measurements such as ATR(14). Market research continues to handle broader synthesis, news, and catalysts. - **Aligned trend and breakout checks.** Updated guidance pairs the latest completed trend window with the preceding breakout baseline and uses a shared observation time for related indicator requests. For example, compare a seven-day change over 42 completed 4h candles with a breakout above the preceding 42 candles' highest high, then request 4h ATR(14). These chat lookups are read-only. Strategy and settings changes retain the proposal and approval flow, and autonomous technical checks still require Technical Analysis to be enabled. A chat result does not replace fresh evidence on a later trading turn. ## 0.9.50 **October 3, 2026** Expanded technical analysis, persistent agent-chat context, and additional execution controls. ### Added - **Agent-scoped chat memory.** When enabled, saved discussion preferences, goals, and user-adopted market theses carry into new conversations with the same agent. Ask the agent to save, correct, or forget a detail. Background processing can also capture durable context from newly completed exchanges. Chat memory remains separate from trading notes and does not change strategy or settings. - **Maximum Open Positions.** Set an optional cap on distinct open symbols on the selected account. Manual positions, positions opened by other agents on that account, and pending entries count toward the cap. Increasing an existing position does not consume another slot. Reductions and closes remain available, and lowering the cap does not force closures. - **Re-entry Cooldown.** Set an optional waiting period, in minutes, after a symbol is fully closed. The rule covers subsequent long and short entries, including after manual closes and liquidations. Partial reductions do not start the timer. Changing the duration applies to the last recorded full close. - **More context in agent chat.** Chat can consult the selected agent's trading notes and request broader market research. Reading trading notes is available independently of chat-memory capture. Position caps and cooldowns govern agent entries; manual trading retains its existing permissions. Both settings accept positive whole numbers, and leaving them unset disables the respective limit. See [Agent settings](/reference/configuration/) and [Chat and instructions](/guides/chat-and-instructions/) for configuration and approval flows. ### Improved - **More powerful technical analysis.** Trading agents with Technical Analysis enabled can access indicator and completed-candle measurements directly, building on their existing TA and research capabilities. Direct tools handle precise numerical checks; market research continues to provide broader analysis, news, and catalyst context. - **Trend and breakout measurements together.** Each candle-statistics request returns the latest trend window, preceding baseline windows, and the latest completed comparison candle. For a seven-day trend plus a 42-candle breakout on 4h candles, the trend window includes the latest candle and the breakout baseline excludes it. The seven-day measurement ends at the latest completed candle, rather than the current moment. - **Larger candle scans.** Check up to **20 symbols per candle-statistics request** when timeframe, window size, window count, and observation time match. Compact tables retain numerical precision, timestamps, coverage, and comparisons, with shared definitions reported once. This does not automatically make every turn scan every market. - **Explicit indicator selection.** Updated guidance helps agents request indicators and periods such as 4h ATR(14), including for volatility-based stop calculations. ATR must be requested explicitly; it is not part of the default indicator set. - **Clearer strategy reviews.** Chat distinguishes evidence recorded under earlier configurations from behavior under updated settings and identifies when relevant history is insufficient to assess a change. ### Fixed - **More reliable candle requests.** Candle-statistics calls now accept the string-encoded symbol arrays some models produce, avoiding unnecessary validation failures. - **Actionable data-quality reporting.** Routine unverified-integrity warnings no longer appear in agent-facing candle results. Actual missing, incomplete, or invalid candles still make affected measurements unavailable, and required unavailable evidence prevents the dependent entry. - **Clearer leverage rejections.** Opening orders require a known exchange maximum leverage. Requests above that limit are rejected before submission with the requested and allowed leverage, instead of a generic leverage-update failure. Missing leverage metadata is reported as unavailable. Closing orders retain their existing path. ### Tool integration changes For callers of `market_candle_statistics`: - Remove the `compare_latest` argument. Responses always include `latest_window`, preceding `windows`, and `comparison_candle`. - Read the full evidence from the compact text tables. Structured content now contains only status, format, and result/error counts; it no longer contains the full `results` array. - Candle statistics accepts up to 20 symbols. The separate indicator tool, `market_technical_analysis`, retains its 10-symbol limit. ## 0.9.49 **October 1, 2026** ### Added - **Public agent positions.** Read-only access to a public agent's current positions, including symbol, side, size, entry and mark prices, leverage, and unrealized profit or loss. Position results support cursor pagination. - **Public performance history.** Read-only chart data for a public agent, with `1h`, `24h`, `7d`, `30d`, and `90d` ranges and validated chart intervals. - **Visibility protection.** These public responses require an explicitly public, available agent and omit private account identifiers, account names, and private activity details. Private, deleted, or unavailable agents return not found. Public API routes: ```text GET /api/v1/public/agents/{agentID}/positions GET /api/v1/public/agents/{agentID}/performance/history ``` These routes require no authentication and return `Cache-Control: no-store`. Existing authenticated routes are unchanged. ## 0.9.48 **October 1, 2026** ### Improved - **Clearer position context for trading agents.** Positions are supplied as labeled records, making fields such as side, size, entry price, margin, and leverage explicit while preserving numerical precision. Unavailable values remain marked as unknown. - **More focused trading memory.** Resolved and other explicitly completed notes are excluded from active trading context while their stored history is retained. Active and ambiguous notes remain available to the agent. - **Less repetition during memory updates.** Trading-memory saves and deletes return concise confirmations and focused correction details, reducing repeated context during a turn.