إنشاء وإدارة أول وكيل لك — الدليل الكامل
جولة في كل شاشة وكل حقل وكل زر في تدفّقَي إنشاء الوكيل وإدارته.
Launching an agent on SorooshX takes about a minute end-to-end. This guide walks you through every screen you will see, what each field controls, what happens on the backend when you submit, and how to manage the agent after it goes live. Bookmark this — it is the reference for everything an owner can touch.
You do not need an email, a password, or KYC. You need a connected wallet and (for your first agent) an invite code.
Before you start
- Connect a wallet at the top of any page (MetaMask, Rabby, WalletConnect — any EVM wallet works).
- Sign in: a single message signature proves you own the address. No funds move.
- Get an invite code from a friend who already uses SorooshX, or open any /ref/ link to be auto-attributed. You only need one — every agent you create after that is unrestricted.
Two ways to create: Prompt mode and Form mode
The top-right toggle on /agents/new switches between two creation paths. They produce the same kind of agent — pick the one that fits how you think.
| Mode | Best for | Tradeoff |
|---|---|---|
| Prompt | Describing the strategy in plain English to an AI assistant. Good for first-time owners. | Slightly slower (one chat turn); requires Smart or higher plan tier. |
| Form | Picking concrete values directly. Good for experienced traders or copying a preset / community strategy. | You make every decision yourself — no AI hand-holding. |
Prompt mode — talk to the assistant
The assistant asks questions, you answer. After 5 fields are filled (name, market, strategy, risk, leverage), the assistant compiles your description into a ready-to-create agent.
What you see on screen
- Progress dots — five dots under the page title. Each fills when the assistant captures another required field.
- Preset shortcut row — three tested presets shown before you start typing. Click one to skip the chat entirely and jump to a pre-filled form.
- Chat composer — large textbox at the bottom. Enter sends; Shift+Enter inserts a newline.
What to type
Be specific about market, side preference, and risk appetite. Examples that work well:
- "Long-only ETH momentum on 1h candles. Conservative — risk 25, 5x leverage."
- "BTC mean-reversion on 15m. Looking for setups when price stretches far from EMA(50). 3x leverage, take quick wins."
- "HYPE breakout on 5m. Both directions. Tight stops because HYPE moves fast."
Form mode — every field explained
Form mode shows every knob the engine accepts. Top to bottom:
Strategy preview card (only when copying)
When you land via a community-strategy "Use this" button, a pasted sse:strat:… config token, or a ?config= URL parameter, you see a brand-coloured highlight card at the very top of the form. It shows the strategy name, the author (linked to their Ideas profile), the description they wrote, and chips for symbol / timeframe / risk / leverage / strategy. Click "Show rules" to see a plain-English summary of the rules that will run. Click the ✕ to clear the strategy and start blank.
Starting preset cards
A grid of 6–7 tested presets at the top of the form. Each preset is a sane starting point aligned with the engine — vol-aware position sizing, ATR-based stops, signal filters all active. Clicking a preset writes its symbol, timeframe, strategy, risk, and leverage into the form below. You can edit any field after.
- SOL Momentum (5m) — trend-following on SOL.
- ETH Steady Trend (1h) — slower, hands-off momentum on ETH.
- BTC Cautious (15m) — lowest-risk mainstream config.
- SOL Breakout (5m) — range-break entries on a vol-heavy token.
- HYPE Vol Hunter (5m) — high-volatility native HL token.
- ETH Mean Reversion (15m) — counter-trend on ETH.
- Start from Scratch — pick everything yourself.
Agent name
Free text, max 80 characters. Shown publicly on the agent card, the chart header, and the share preview. Pick something descriptive — "ETH 1h Trend Hunter" beats "Agent 7". The URL gets a stable random suffix so renaming later never breaks indexed links or shares.
Description (bio)
Free text, max 2000 characters. Public. Use this to tell other traders what your agent does, when it works, and when it does not. Good descriptions get followers; cryptic ones do not.
Market
The HL perp the agent will trade. One agent = one market — if you want to trade both BTC and ETH, create two agents. Pick a market with enough volume that fills and exits are realistic; majors (BTC, ETH, SOL) are forgiving, thinner markets require tighter risk settings.
Candle resolution (timeframe)
How often the engine reads new data and reevaluates. Available: 1m, 5m, 15m, 1h, 4h, 1d.
| Timeframe | Decisions per day | Best for |
|---|---|---|
| 1m | ~1440 | Scalping volatile coins (HYPE, SOL). High activity, high fee drag. |
| 5m | ~288 | Active intraday on majors and mid-cap perps. Most balanced default. |
| 15m | ~96 | Less noisy, fewer false signals. Good for mean reversion. |
| 1h | ~24 | Trend-following with daily checks. Lowest-stress option. |
| 4h | ~6 | Swing trading. Few trades, larger moves. |
| 1d | ~1 | Long-term positioning. Mostly for testing macro thesis. |
Max leverage
Upper bound on the position size relative to your capital. 1 = no leverage; 50 = the platform maximum. The engine will not exceed this even if a training prompt asks for more.
Trading strategy
Three built-in archetypes that decide the agent's base behaviour:
| Strategy | What it does | Likes |
|---|---|---|
| Momentum | Buys strength, sells weakness. Trades with the trend. | Markets with clear directional drives. |
| Mean Reversion | Fades extension. Positions for a return to fair value. | Range-bound markets, quieter regimes. |
| Breakout | Waits at range edges. Enters when price clears a level. | News, level retests, volatility expansions. |
A fourth option, Custom (DSL), only appears when you copy a community strategy or paste a strategy-config token — it runs the exact rules the author wrote.
Training model (LLM tier)
Which AI tier the Train Agent chat uses to interpret your training prompts. Does not affect trading decisions — only the quality of the AI's understanding when you talk to it.
- Quick — fastest, cheapest. Sufficient for most prompts.
- Smart — better at multi-step adjustments.
- Deep — slowest, most thorough. Use when training feels off.
- Mini — placeholder for the smallest available model.
Risk level (the slider)
A 0–100 slider that controls how aggressively the engine sizes positions. The label updates as you drag: Conservative (under 33), Balanced (33–66), Aggressive (67+). Internally this maps to volatility-aware sizing — lower numbers mean smaller positions when the market is choppy, larger numbers mean the agent leans into volatility.
Submit and the invite-code modal
Clicking Create on either mode submits the form. If your wallet has not been attributed to a referrer, an Invite Code modal pops up:
- Enter any valid code (case-insensitive). The platform binds your wallet to that referrer permanently.
- You only see this modal once. Every future agent skips it.
- No skip button — codes are mandatory. Open any /ref/<code> link if you don't have one.
What happens after you submit
- The agent is created in Paper mode with a 10,000 USDC demo bankroll.
- You are redirected to /agents/<your-slug>.
- The agent does NOT start trading automatically — you start a paper run from the right-side panel.
- A retire button appears in Manage; the agent can be soft-deleted any time without losing its history or URL.
The Manage Agent modal — every control
Click Manage on any agent you own to open this modal. Changes here apply on the next evaluation tick.
Name + avatar
Edit the display name (max 80 chars) and the avatar URL. Leave avatar blank for an auto-generated identicon. The URL of the agent NEVER changes when you rename — old shared links keep working.
Bio
Public description, same field as on Create. Edit anytime.
Market
Change which HL perp the agent trades. Saving is blocked while the agent is running in paper mode — pause it first, otherwise the open position would be stranded on the old market.
Candle resolution
Switch timeframe (1m → 1d). Same constraint as Market — pause first if there's an open position. The new timeframe takes effect on the next tick.
Strategy label
Free-text tag shown on the agent card. Cosmetic only — does NOT change trading behaviour. Use it to give your strategy a memorable name like "Funding Carry v2".
Trading strategy + AI model
Switch between momentum / mean-reversion / breakout, or change the AI tier used by Train Agent. Switching strategy mid-run does not close existing positions.
Risk + Leverage sliders
Same controls as on Create. Adjust freely — the new caps apply to new entries.
Public / Private toggle
Public agents appear in the directory and on community pages. Private agents are only visible to you. You can flip this at any time without affecting the agent's state.
Start / Pause button
Pause stops the engine from evaluating ticks. Open positions stay open — the engine will not enter or exit until you Start again. Useful before major news or while you reconfigure.
Retire (soft-delete)
A two-step destructive action at the bottom of the modal. Retiring an agent:
- Stops trading immediately and removes the agent from public lists.
- Keeps the URL alive — anyone with the link still sees the full page with a "This agent has been retired" banner. Historical performance and JSON-LD stay intact for SEO.
- Disables Run / Backtest / Train / Share / Manage on the retired agent.
The Run form — starting a paper run
On the agent detail page, the right-side panel is the Run form. It controls a personal paper-trading instance of the agent for your wallet.
- Paper / Live toggle — Live is not enabled yet; agents only paper-trade. Live needs additional safeguards before we ship it.
- Available balance — your wallet's demo USDC bankroll (starts at 10,000).
- Invested amount — how much demo capital this run gets. Max button uses your full available balance.
- Risk control slider — overrides the agent's default risk for THIS run only. Does not change the agent's saved risk level.
- Run Paper Agent button — starts the run. Stop Agent (when active) closes the open position and credits equity back to your demo balance.
Owner action row (under the run controls)
- Backtest — opens the backtest modal. See the dedicated backtesting article for details.
- Manage — opens the Manage Agent modal described above.
- Share — opens the Share modal with two tabs: Publish to community library, or Copy as code (a sse:strat:… token others can paste into their own /agents/new).
- Train Agent — opens the chat at the bottom of the page. Plain-English tuning that the engine applies safely.