Deepfriday
Try it for free

//The first configurable multi-agent investment platform

Build your own AI investment firm.

Design a team of AI agents (analysts, reviewers and a decision-maker) and put them to work on more than 1,000 expert investment articles a day from over 20 sources. They don't guess from what they absorbed in training: every pitch, objection and position is argued from theses that experienced investors wrote.

Try it for freeSee a run

Free to start. No card required. Your first runs are on us.

Expert research sources
20+
Investment articles analysed a day
1,000+
Layers you design yourself
8
AI model providers, chosen per agent
5

Paper portfolio. Nothing here places a real trade.

01Expert research

Agents that think with the experts, not from memory.

A language model left alone answers from whatever it absorbed in training: dated, generic and impossible to check. Deepfriday's agents never work that way. Every run starts from investment theses and ideas written by human experts, and every finding an agent makes is tied back to the analysis it came from.

How it works

  1. 01

    Expert theses arrive

    More than 1,000 investment articles a day from over 20 expert sources, captured around the clock.

  2. 02

    Agents read the arguments

    Each agent weighs real analysts' reasoning (the thesis, the catalyst, the risk) before forming a view of its own.

  3. 03

    Every idea cites its research

    Positions carry the analysis behind them, so you can check what an agent read instead of trusting what it says.

Go to

Human theses1,000+ a dayTraceable

02Scale

Read in an hour what no analyst could read in a month.

No human desk can read a thousand expert articles a day, check every ticker and remember yesterday's view. Your agents do it on every run. Each article is distilled into a structured note before any agent sees it, so the desk spends its effort on judgement, not on reading.

Assets
Every company, ETF and index an article names.
Intent
Long, short or watch: what the author actually argues.
Symbols
Each ticker confirmed against live market data.
Rolling view
How expert opinion on a name is shifting.

Go to

Structured notesValidated tickers

03The draft

Every stock gets one owner, and all the evidence.

Research is dealt by stock, not by article. Each analyst receives a private shelf of names, each one bundled with every note written about it. Nobody pitches a company having read a fraction of its coverage, and no two agents chase the same ticker.

How it works

  1. 01

    Grouped by stock

    The day's notes are regrouped into one lot per ticker, freshest first.

  2. 02

    Dealt fairly

    Lots are spread across analysts, balanced by depth of coverage, and rotate so every seat sees every shelf.

  3. 03

    Claimed with conviction

    Analysts claim the names they can argue. A dropped name gets a second chance, with the reason attached.

Go to

One lot per tickerNo duplicates

04Your firm

Your firm. Your org chart. Your rules.

Nothing about the team is fixed. Design it the way you would staff your own investment firm: how many layers it has, what each one is for, how many agents sit in each, which model every agent runs on and what kind of investor it is. A private investment firm, staffed entirely by AI agents.

Layers
Two to eight, each with a name and a purpose you write.
Headcount
As many analysts or reviewers per layer as the work needs.
Mandate
Your strategy, in your words, shapes every prompt on the desk.
Models
A model per agent from five providers, mixed freely.
Personas
Value hunter, sceptic, momentum trader: give every seat a temperament.
Risk & horizon
Risk profile, position limits, time horizon and run cadence.

Go to

Custom layersAny model

05The debate

Ideas earn their place by argument.

Every pitch is challenged on the figures before anyone sizes it. Reviewers cite valuation, cash runway and the next report, the agent below defends its case, and a verdict is recorded either way. A weak idea has to survive every layer you built before it reaches your book.

How it works

  1. 01

    Pitch

    An analyst makes the case, citing the expert research and the fundamentals.

  2. 02

    Challenge and rebuttal

    A reviewer raises objections. The author answers them with evidence, or concedes.

  3. 03

    Verdict

    Each layer rules on every idea it is handed. Nothing disappears without a recorded reason.

Go to

Full transcriptRecorded verdicts

06The book

Positions you can audit, on a portfolio you control.

The deciding layer sizes the surviving ideas against your risk profile and limits, then proposes concrete moves. Apply them with one click, or let the team allocate on its own. It is a paper portfolio, and every decision is explained.

Proposal
Sized positions, each with conviction, thesis and evidence.
Apply
One click, or automatic allocation at the end of every run.
Plans
Targets, stops and review dates, checked on every run.
Digest
An email summary when each scheduled run finishes.

Go to

Paper portfolioEmail digest

A run, as the desk files it

Sample output. Figures are illustrative, not a track record.

  1. deskRun started · 24h window · 1,140 expert articles in the corpus.
  2. desk612 stock lots dealt across 6 analysts.
  3. analystNVDA LONG: 14 expert notes agree data-centre demand outruns supply through next year, and margins are holding.
  4. reviewerValuation already prices that in. What happens if hyperscaler capex slows after the next report?
  5. analystEarnings growth still outpaces the multiple, and the three newest notes raise capex guidance rather than cut it.
  6. reviewerAccepted. The objection is material, not fatal: conviction lowered from 9 to 8.
  7. leadNVDA sized at 12% of the book, funded from cash, within the 15% position limit.
  8. deskRun complete · 3 ideas decided · 1 exit proposed.

??Questions

Before you start.

01Is it free to try?

Yes. Every new account starts with free credits, enough for its first runs, and no card is asked for. When they run out you can buy a credit pack, and you only ever pay for what your team actually consumes.

02Does this trade real money?

No. Every portfolio in Deepfriday is simulated. The desk proposes positions and you apply them to a paper book. Nothing here places a real trade, and no broker is ever connected.

03How is this different from asking a chatbot about stocks?

A chatbot answers from what it memorised in training, which is out of date and cannot be checked. Deepfriday's agents read more than 1,000 expert investment articles a day and base every finding on that research. Several agents then argue over each idea before it is sized, and you can read the whole exchange.

04Where does the research come from?

From over 20 expert investment sources, captured continuously: more than 1,000 articles a day. Each one is summarised into a structured note (the assets it names, the position its author argues, a rolling view), and every ticker is checked against live market data before any agent reads it.

05How much of the team can I customise?

All of it. You decide how many layers the team has and what each one does, how many agents sit in each, every agent's prompt, persona and model, and the team's strategy, risk profile, position limits, time horizon and schedule.

06Can I choose which AI models my agents use?

Yes. Every agent takes its own model, drawn from five providers, and you can change any of them between runs. A credits meter reports what each run consumed.

07Can I see why an idea was rejected?

Yes. The board records the whole argument, including the pitches that were thrown out and the objection that killed them. Nothing is dropped silently.

08How often does my team run?

As often as you choose, from twice daily to weekly, or only when you press Run. Scheduled runs email you a summary when they finish.

09What does a run cost me?

A run consumes credits, and one credit is a fixed unit of model usage. The sidebar reports your month to date, and every model's rate is listed in the rate table.

Open your firm tonight.

Write a strategy, design the team, and let its first run read the day's expert research while you sleep.

Try it for freeSign in

Free to start. No card required. Your first runs are on us.

Click an agent to meet it