Features

See it before
you download it.

Every screenshot and clip on this page is London5 itself, running on a MacBook Pro with an M5 Pro chip. The prices and the news are live, and the AI answering is a model running on that same Mac.

Try it freeTake the tour

Conversation

Ask in plain words.
Get the chart, not a paragraph.

Talk to London5 the way you would to a colleague. Pip, your sidekick, reads what it needs, then pushes the chart, the table or the source into the conversation, where you can check it.

Try it here. A working model of the app, running in your browser on a simulated market: pick a suggestion or type, hover a chart, change its range, pin it to your board, or switch to Cockpit at the top.

London5 conversation: Apple, Microsoft and NVIDIA compared over six months, with the chart pushed into the answer
“Compare Apple, Microsoft and NVIDIA over the last six months.” The chart is live and pinnable; the answer took 52 s on the Mac's own model.

News & Feeds

Every story that moves
your markets, sorted.

London5 reads the wires and the filings as they land, tags each story with the markets it touches and how it leans, and keeps the ones about your symbols on top. Tap a number to see what each part does.

London5 Markets, Feeds, Episodes view
Markets → Feeds → Episodes, 3 October 2026.

    Dashboards

    Many charts.
    One glance.

    Build boards of charts and tiles for the markets you follow, and switch between them. Every tile stays live while the market is open.

    London5 Dashboards, Markets board of indexes
    The world's indexes, crypto and rates on one board.
    London5 Board with pinned charts and a suggested watch
    Your board: pinned charts, the news on your symbols, and a watch London5 suggests pinning.
    London5 Markets overview, currencies
    One market at a time: here currencies, with the session and its movers.
    London5 Markets sentiment
    Market mood beyond the price: fear and greed, sector by sector.

    Research New version coming

    Research that shows
    where every finding comes from.

    Ask a question about a company and London5 reads the reports, the calls and the news, and writes up what it found, with each finding linked to its source.

    London5 Research reading documents about ASML
    A research run on ASML, reading 43 documents. These are the current screens; they will be replaced when the new Research ships.

    Simulator

    Six strategies. Real past prices.
    All on paper.

    Before you put a strategy to work, run it over real past prices with paper money and see what it would have done: the return, the fees, the worst drop it would have made you sit through, and how it compares with simply holding.2

    Simulation verdict: NVIDIA, just hold

    NVIDIA, just hold

    Buy once, hold to the end.

    +15.3%on paper
    Simulation verdict: Microsoft trend follower

    Microsoft trend follower

    Buy when the 10-day average rises through the 30-day; sell when it falls back.

    +4.2%8.6 points behind holding
    Simulation verdict: S&P 500, a little every day

    S&P 500, a little every day

    Buy $100 of the S&P 500 fund on a daily schedule.

    +1.7%on paper
    Simulation verdict: Tesla breakouts

    Tesla breakouts

    Buy when Tesla closes above its 15-day high; exit on a volatility stop.

    +1.1%4.2 points ahead of holding
    Simulation verdict: Apple, buy the dips

    Apple, buy the dips

    Buy when Apple closes under its lower Bollinger band.

    0.0%the rule never fired
    Playbook run: Core three, three sleeves

    Core three

    A playbook: S&P 500 50%, Nasdaq 100 30%, gold 20%, held.

    +2.0%still replaying
    A simulation replaying its days, sped up.
    StrategyReturnWorst dropOrdersHolding the whole $10,000
    NVIDIA, just hold+15.3%10.0%3+15.6%
    Microsoft trend follower+4.2%2.7%4+33.5%
    S&P 500, a little every day+1.7%3.0%14+2.3%
    Tesla breakouts+1.1%2.9%2−16.6%
    Apple, buy the dips0.0%0.0%0+13.1%
    Core three+2.0%—3—

    Paper capital on past daily prices, 1 July – 30 September 2026, $10,000 each. A result on past prices doesn't tell you what comes next, and it isn't a recommendation.

    Watches and alerts

    Set it,
    and get on with your day.

    Tell London5 what to watch for, a price, a move, a story, and it keeps watching while your Mac stays on. When something happens it tells you, with what changed next to it. A watch only tells you; it never acts.

    London5 Watchtower with a watch that fired
    Watchtower: your watches, and the sighting when one fires.
    A watch firing, and the board telling you.

    // under the hood

    The AI runs on this Mac.
    Here are its numbers.

    For the curious: London5 measures every model call it makes, how many tokens, how fast, from which agent, and shows it under Settings → Intelligence → Usage. These are the numbers from the session that made this page.

    Machine
    MacBook Pro · Apple M5 Pro
    Memory
    48 GB unified · 38 GB for the GPU
    Model
    Qwen3.8 27B · 3-bit · 11 GB
    Licence
    Apache 2.0, open weights
    Context
    16,384 tokens · 4 slots
    Writing speed
    17.8 tokens/s
    Reading speed
    220–360 tokens/s
    This session
    1,131 requests · 892,866 tokens
    AI bill
    €0.00

    Measured on 3 October 2026 while the news classifier, a research run and up to eight simulations shared the model. One chat turn read 10,516 tokens of context in 36 s and wrote its answer at 10.5 tokens/s.

    London5 Settings, Intelligence, Usage
    Settings → Intelligence → Usage, during the session that made this page.
    London5 usage by agent
    Which agent used the model, how much, and how fast.
    London5 local model memory
    What each model would cost this Mac to run: weights plus context.
    llama-server · 127.0.0.1 · this session
    # llama-server, started by London5 · Qwen3.8 27B GSQ-RCO IQ3_S
  - MTL0    : Apple M5 Pro (38338 MiB, 38338 MiB free)
llama_context: n_ctx_seq (16384) < n_ctx_train (262144) -- the full capacity of the model will not be utilized
srv    load_model: initializing slots, n_slots = 4
srv  llama_server: model loaded
srv  llama_server: server is listening on http://127.0.0.1:19554
# 12:57 · chat: “Compare Apple, Microsoft and NVIDIA over the last six months.”
slot print_timing: id  1 | task 5073 | prompt eval time =   35996.44 ms / 10516 tokens (    3.42 ms per token,   292.14 tokens per second)
slot print_timing: id  1 | task 5073 |        eval time =    5903.65 ms /    62 tokens (   95.22 ms per token,    10.50 tokens per second)
slot print_timing: id  1 | task 5073 |       total time =   41900.08 ms / 10578 tokens
# 13:05 · news: classifying a headline
slot print_timing: id  3 | task 11618 | prompt eval time =     748.55 ms /   189 tokens (    3.96 ms per token,   252.49 tokens per second)
slot print_timing: id  3 | task 11618 |        eval time =    2753.95 ms /    52 tokens (   52.96 ms per token,    18.88 tokens per second)
slot print_timing: id  3 | task 11618 |       total time =    3502.50 ms /   241 tokens
# 13:05 · news: a longer article
slot print_timing: id  2 | task 11731 | prompt eval time =    3265.84 ms /  1186 tokens (    2.75 ms per token,   363.15 tokens per second)
slot print_timing: id  2 | task 11731 |        eval time =     486.68 ms /    10 tokens (   48.67 ms per token,    20.55 tokens per second)
slot print_timing: id  2 | task 11731 |       total time =    3752.52 ms /  1196 tokens

    Try LONDON5 free for 60 days1

    Everything above, on your own Mac. Create your London5 account with your email, and we’ll email you a code to download the app. Requires macOS 14 Sonoma or later; Apple silicon recommended.

    Try it free for macOS
    1. Cloud AI and data providers you choose may charge you separately. London5 charges no per-trade or performance fees.
    2. Simulations run on paper capital. A result on past prices doesn't tell you what comes next, and it isn't a recommendation.