# Show HN: Basis – A free crypto terminal without paywalls > Source: > Published: 2026-08-19 10:21:02+00:00 # A crypto terminal with an agent that uses it. Not a chatbot beside a chart. 67 tools wired into the real terminal — it loads the data, computes the indicators, draws on your chart, screens the market, writes strategies and sets the alerts. You bring the API key; we take nothing. [Open the terminal](/?app=1)Charts work before you sign up. Always will. Mark the level BTC keeps reacting to, tell me if the derivatives agree, and set an alert if it breaks. 61,000 is drawn and the alert is live. Funding is mildly negative and open interest fell into the last test, so the derivatives lean with the level rather than against it — but the 4h RSI at 38 is not oversold, so this is support holding, not a reversal signal. The agent ## Say what you want. Watch the terminal do it. Most tools bolt a chat window onto a chart and call it AI. This one hands the assistant the terminal itself — one tool per thing the app can do — so the answer arrives as work already done, on your chart, rather than as a paragraph telling you to go and do it. ### Reads the chart Loads candles at any timeframe, computes any of the 93 indicators, and reads derivatives and venue-tagged order flow. ### Draws on it Places any of the 68 drawing tools at real levels — trendlines, zones, Fibonacci, measured moves. ### Screens the market Builds a screen from your description, runs it across the universe, and backtests the screen itself. ### Writes strategies Programs in a sandbox, searches the parameters, and publishes the ones that survive a held-out window. ### Watches for you Sets alerts on price, on indicator conditions, and on a screen coming true — then tells you when it does. ### Uses your key Anthropic, OpenAI, Google, xAI or OpenRouter. The key is stored in your browser and never reaches a server of ours. Some of what it can call The price ## There is no plan to compare. There is one, and it is all of it. The things other terminals meter — how many indicators on a chart, how many saved layouts, how many alerts, whether you may open a second chart at all — are not metered here. Not generously limited. Not limited. ### 93 indicators On as many charts as you like, as many at a time as you like. ### Unlimited alerts Price, indicator and screener conditions, delivered to the browser. ### Unlimited layouts Save every workspace you build. Multi-chart included, not upsold. ### 20 chart types Footprint, market profile, volume profile, Renko, Kagi, Point & Figure and the rest. Range bars too, as a timeframe. An account is free too, and optional. Charts, indicators, drawings and backtests all work signed out — an account exists only so your saved work follows you to another device. What is underneath ## Built for the parts of crypto that general charting tools skip. ### 57 functions, your own indicators Write an expression, plot it on the price pane or its own, and it recomputes as the chart updates — a live series, not a stored screenshot. ### A spreadsheet over live markets Columns of market fields, indicators and your own formulas across a universe of symbols. Sort it, filter it, push a column to the chart, export it. ### A properly synchronized book L2 depth from a sequence-checked snapshot, exact aggregate trades, a footprint, and a liquidity heatmap that starts when the book connects. ### Funding, OI, positioning, liquidations On the same time axis as price — the layer a multi-asset charting tool structurally cannot carry, because it is not built for one market. ### 33 built in, or write your own A sandboxed program in your browser with causal helpers, parameter search against a held-out window, and walk-forward folds. ### Which of your setups actually pays Expectancy in R after fees, cut by setup, symbol, tag and strategy — including whether following your own plan made you money. And it will not flatter you ## The number a backtest gives you is the easiest number in the world to fake. Search enough parameter combinations and one of them scores beautifully on pure noise. Basis reports how many were tried, what the best of that many would have scored on random data, and whether the ranking survived a window it never saw. Best of 219 configurations tried. A Sharpe of 2.63 is what searching this many configurations produces from pure noise (4.30), so this result is not evidence of an edge. Rankings did not carry into the held-out period — the search fitted noise. Actual output from the strategy search Every rule you write clears a causality gate first: the program is re-run over historical prefixes and the results compared, so one that quietly read the future cannot be scored at all. Not a warning — a refusal. ## Open it and put something on the chart. No sign-up, no card, nothing to cancel. You will know inside a minute whether this is the terminal you wanted. [Open the terminal](/?app=1)