Start with the decision your scan should support
A useful scan starts with a focused question. You might want a morning comparison of your crypto watchlist, a review of new company filings or a weekly check on a market thesis. Explain what you need to understand and what would make a finding worth your attention.
In Premiss, your agent works from that brief and saves useful findings for later runs. Include your timeframe, assumptions and preferred report format so it can keep the research relevant.
Choose the markets and the evidence
Name a manageable set of stocks or crypto markets. Specify the exchange or trading pair where it matters, and the interval you want compared. A daily close and an unfinished intraday candle answer different questions. Ask the agent to identify the period behind each observation.
Combine price history with sources that fit your question. Company research may call for SEC filings and supporting web research. A crypto watchlist review may begin with historical price bars, then investigate the context behind a move. Ask for links, source dates and any gaps in the available evidence.
- Market scope: selected symbols, trading pairs and any exclusions.
- Comparison: a defined period, baseline or previous research finding.
- Evidence: the sources to prioritize and the questions they should answer.
Write a task your agent can repeat
Example brief: “At 09:00 Europe/Stockholm every weekday, review BTC/USD, ETH/USD and SOL/USD on the same selected exchange. Use the latest completed daily candle. Compare each close with the previous close and its close seven days earlier. State the exchange, interval and candle timestamp. Investigate context for the largest moves, favor original announcements and include source links. Say when information is missing. Send the report to my agent chat. Keep this task in research mode.”
Review the symbols and source availability before saving this example. Ask for the same report structure each time:
- Snapshot: markets, comparison periods and the time of the data used.
- Findings: price comparisons and relevant context retrieved during this run.
- Evidence: source links and publication dates, with interpretation clearly identified.
- Follow-up: unanswered questions and changes to the next research brief.
Set the schedule and the report destination
Choose a frequency and timezone, then save the recurring task. Match the schedule to the information you need: a daily price review, a weekday filing check or a weekly thesis update. A schedule determines when research runs; the freshness of each finding still depends on its source.
Send the report to agent chat for follow-up questions, to Desk for a shared research record, or to a connected Discord channel. A filing review and a crypto scan can have different schedules and destinations.
Choose feedback, review or automatic action
A scan can finish with a briefing, propose a trade for your review or lead to an automatic trade. Choose the outcome in the task. If you enable trading, specify the markets, position limits and trading connection.
Describe the conditions that require your review. The agent can weigh new research before deciding to act, or apply a strategy with fixed entry and exit rules. In either case, its findings and the actions taken should be available for you to inspect.
Improve the scan from the reports you receive
After a few runs, check whether the findings are useful, the sources are relevant and the brief is precise enough. Ask the agent to explain an unsupported conclusion, narrow an overly broad watchlist or carry a correction into future research. Saved findings and feedback give subsequent runs a clearer starting point.
If a scan suggests a repeatable trading idea, define and backtest its rules, then observe it in paper trading before considering automatic execution. Evaluate the behavior and assumptions as well as the result. Historical and simulated results do not guarantee future performance.