Strategies — Trade While You Sleep

Strategies let you automate any instruction — no matter how complex — and have Fere run it on repeat. Your strategies run on autopilot — buying, rebalancing, and watching the market while you do literally anything else. Describe what you want once, and Fere handles the rest every single cycle.

​

What Are Strategies?

A strategy is an automated instruction that runs on repeat. You describe it in the chat at /app, tell Fere how often to run it, and sit back. Every cycle, the agent goes through the full reasoning process — scanning, analyzing, deciding, and (if conditions are met) trading. And here’s what makes it powerful: Fere remembers.

​

Gets Smarter Over Time

Every time a strategy runs, it carries forward what it found before. By day 7 of an hourly scan, your agent has a rich picture of every token it analyzed, every score it gave, every trade it made or skipped. It can spot things a stateless bot never would — like a token that scored 85 three days ago, dropped to 60 yesterday, and just came back to 82. That’s a recovery signal, and Fere catches it because it was watching.

​

How to Create a Strategy

Go to the main chat at /app and describe what you want to automate. The AI sets it up and it appears in /strategies. A few examples to get you started:

​

Example: Hourly Token Scout

Runs Every Hour “Go to Base chain and find tokens between $10M-$20M market cap. Analyze them on socials and technical analysis and give them a score based on performance. Buy a token if it scores 90/100.”

Every hour, Fere will: scan the chain for candidates, run social and TA analysis on each, score them, and only execute a trade if the quality bar is met. Open the strategy detail page after a run to read the executive summary, inspect past runs in the sidebar, and review any trades or files the agent produced.

​

The Strategies Page

Click Strategies in the top navigation bar. Here’s what you’ll see:

• Active / Paused tabs — The page opens on Active. Switch to Paused to see anything you’ve stopped, with counts on each tab so you always know where you stand.
• Strategy cards — Each card shows your original instruction, how often it runs, when it runs next (or last ran if paused), and how many times it’s executed so far.
• Start a Strategy — At the bottom of the page, pre-built templates let you launch a strategy in one click without typing anything.

​

Start a Strategy

The Start a Strategy section inside the Strategies page has pre-built templates that get you running in one click. Pick one, fill in a few parameters, and Fere creates the full automated strategy for you.

​

Strategy Detail Page

Click any strategy card to open its detail page at /strategies/[id]. This is your command center for one strategy — what it has done, what it traded, what it produced, and what you want it to do next.

​

Executive briefing

At the top of the page, a briefing card shows:

• Strategy name and description — the label Fere assigned plus your original instruction.
• Health and headline — whether the strategy is on track, degraded, blocked, paused, or research-only, plus a one-line verdict.
• Key metrics — run success rate, trade outcomes, and next scheduled run.
• Since last update — bullet changes since the previous briefing (incremental window).
• Top positions — largest holdings when portfolio data is available (hidden for research-only agents).
• Narrative sections — at-a-glance, recent performance, and since-last-update prose grounded in verified run and trade data.

A compatibility markdown summary is still stored for exports and older clients. The briefing refreshes after each completed run. If a run just finished, you may briefly see “Updating briefing after the latest run…” while new content is generated.

​

Recommendations

Recommended actions sit in the same briefing card, directly under the metrics and positions. They are separate from the narrative briefing: short, actionable cards tuned for what to do next, not a prose recap of the last run.

​

What you see on each card

Each recommendation includes:

• Priority — Now (urgent), Soon, or Later, so you know what to tackle first.
• Category — Execution, Risk, Logic, Config, or Research.
• Title and why — a plain-language summary and the reason Fere surfaced it.
• Steps — one to five direct imperatives (for example, “Extend the schedule for 3 more runs” or “Set a take profit at 5% above entry”). Steps are written for you as the owner, not as developer instructions.
• Evidence (when shown) — pointers to the run, trade, signal, or script behavior that supported the suggestion.

Fere shows up to five recommendations at a time. If nothing new is warranted, the section may be empty.

​

What you can do

While a new pass is running after a completed run, you may see “Updating recommendations after the latest run…” with no cards yet. That is normal; reload or wait a moment for the job to finish.

​

How recommendations are generated (high level)

Recommendations are produced by a dedicated background job that runs in parallel with the executive briefing — not copied from the briefing text.

1. Gather evidence — Fere assembles a structured snapshot of your strategy: recent and older scheduled runs, trade outcomes, holdings, error patterns (with extra weight on recent failures), your active strategy text, saved workspace scripts when present, optimizer metrics if enabled, and your past Strategy chat follow-ups.
2. Detect signals — Rule-based checks flag themes worth addressing (for example, repeated run failures, stale errors that are already resolved, or script-specific trading context on Hyperliquid or Polymarket).
3. Generate cards — An AI model reads only that evidence and outputs structured recommendations. Each card must cite supporting evidence; invented balances, prices, or trade counts are not allowed.
4. Sanitize for product context — Post-processing drops generic coding advice (logging, refactoring, “edit your .py file”). Steps stay in owner language: schedule changes, risk guardrails, notifications, and workspace updates via Fere — aligned with how strategies actually run (saved scripts and strategy_builder, not manual IDE edits).
5. Respect your choices — Declined recommendations are stored so refresh passes do not suggest the same intent again.

The executive briefing and recommendations share the same trigger (after a completed run) but serve different jobs: the briefing explains what happened; recommendations suggest what to do next.

​

Strategy chat

Below the summary is Strategy chat — your follow-up conversation with this deployed strategy’s workspace. This is the right place to change what runs on the schedule (not the original /app thread). Strategy chat is also separate from scheduled runs themselves:

• Ask clarifying questions (“Why did you skip the last trade?”)
• Request changes (“Raise the score threshold to 90 and cap buys at $25”)
• Pause or stop the strategy in plain language

Older messages load on scroll (newest at the bottom), so long-running strategies stay fast even after dozens of runs and follow-ups.

​

Run status bar and input

A sticky bar at the bottom of the page shows:

• Run count — how many scheduled cycles have completed.
• Next run — when the strategy fires again (hidden while paused).

Just below that, the input bar accepts new instructions. Type @ to mention another active or paused strategy when you want to reference or coordinate across automations.

​

Right sidebar — Past Runs, Transactions, Artifacts

On desktop, a sidebar on the right replaces the generic Files panel while you are on a strategy detail page. It stays visible for that strategy. If you open the wallet drawer, the sidebar shifts left — it is never hidden behind the wallet. Three collapsible sections stack vertically. Drag the dividers to resize them, and your layout is remembered per strategy: On mobile, tap the floating Strategy button to open a bottom sheet with Runs, Txns, and Files tabs containing the same data.

​

More Strategy Ideas

​

Strategy system capabilities (fere_tools)

Scheduled strategies run Python in a sandbox with fere_tools — a synchronous SDK aligned with Fere’s gateway tools (market data, analysis, wallets, on-chain writes, Hyperliquid, and Polymarket). The full catalog (69 tools, invoke patterns, and examples) lives on a dedicated page so crawlers and readers can load Strategies without a multi-megabyte HTML document.