Tessera is a declarative, composable format for market strategies. The same document you backtest drives live orders — and it runs on your machine, against data you own.
// status: design, in the open. no engine yet — the format, specs and roadmap are public, and we want the first devs, traders and agents in the room.
Testing a market idea shouldn't mean building an engine first.
Three obstacles stand between an idea and an answer — and the second is where most people give up.
Four things, in dependency order.
Declarative. Composed from a small closed algebra over an open vocabulary — a new idea is a new expression, not a new function in someone's codebase. Validated as a whole before anything runs, with errors precise enough for a machine to correct itself.
Runs on your machine, against your data. Same inputs, same version, same numbers — on anyone's hardware. Coverage is reported alongside results; look-ahead is prevented structurally, not by convention.
The document you backtested drives real orders — through a separate runtime, never the backtester's assumptions. A risk gate the strategy can't bypass, idempotent placement, reconciliation, a kill switch, mandatory paper mode.
Vendor files or a broker/vendor API — Zerodha, Upstox, IBKR, Alpaca, TrueData — behind one capability model. Ticks are stored; bars are derived. Convert locally; the data never reaches us.
A whole strategy — and you can move it.
A hedged short strangle: sell an out-of-the-money call and put, buy a wing beyond each, exit on profit, loss, or expiry. Drag the controls — the document and the expiry payoff update live. No engine code; it validates as one document, and an agent can author it from the schema.
Every value carries a unit, so a nonsensical selector is a validation error with a field path — not a six-hour surprise. Grab the full sample or the JSON Schema.
The choices that shape everything.
One thin slice, end to end — then widen.
The risk isn't difficulty; it's never finishing anything end to end. So v0 is one of everything, proven end to end before any axis widens.
We're designing in the open, and we want you early.
There's a public decision log, a per-part spec for every piece of v0, and a tracked roadmap. Traders who want to test ideas without building an engine, quants who want a shared reproducible environment, engineers who want to own a subsystem, and people building agents that read and write strategies — this is the moment to shape it. Early access includes the design docs and the specs.
Be in the room.
Leave an email and we'll bring you in as the pieces land — and reach out if you want to help build one.