TRA

[TRA by Bijective Labs]

TRA

Autonomous trading agents with deterministic execution. The agents propose; the system decides.

[CI] [PyPI] [Python versions] [License: MIT] [Status: alpha]

[Managed with uv] [Linted and formatted with Ruff] [Type checked with mypy] [Venues through ccxt] [Tested on Linux and Windows]

[Website] [Documentation] [Insights] [Changelog] [Security policy]

TRA (trading research agents) is published as bitra on PyPI and runs as tra. Before 0.10.1 the package was called trading-research-agents, and the same package is still published under that name.

Language models are good at reading a market and saying what they think. They are bad at arithmetic, they don't know what a trade costs, and they will happily overrule a limit you set. This project uses them for the first part only. The system itself, fixed rules with no model involved, fetches the data and works out every figure. The agents read those figures and give an opinion. Then the system prices the trade, applies your limits, sizes it, and either writes it down or sends one order with the stop attached. Every decision and every refusal is kept on disk with its reason, so you can check afterwards what happened and why.

It's an early release, so expect settings to change between versions. What works today: linear perpetuals, spot pairs (stock perpetuals and tokenized stocks among them) and prediction market outcomes through one order path, a universe of perpetuals found and measured by the system and ordered by the room its own paths expect, the full autonomous cycle on live data, a structure lead that proposes, three readings of your own series computed by the system, up to three reviewers, stops and targets the lead names inside bounds you set, the gates that decide, sizing from your own rules, real orders on a venue's practice environment behind a mandate you sign and a signed risk gate, and two auditors that read the records. Under it, a market data plane records order flow, the order book, liquidations, derivatives and news minute by minute into a verified archive, and an arena replays any policy over that archive beside baselines built to be beaten, net of the desk's own costs.

Contents

• What it solves
• How it is built
• Get started
• What you see, in order
• How one cycle decides
• Venues
• Models and providers
• Toward real orders
• Your money, your orders, your positions
• Watching a desk, and what is kept
• Command line
• What it guarantees, and what is not built yet
• Where it is going
• Documentation

What it solves

None of these is hypothetical. Why it is built this way lists the documented failures behind each one, with their sources.

How it is built

Three kinds of work, kept apart because they fail differently. The system is in control; Jev, TypeSafe's typed judge, is the fast System One model that perceives and verifies; the reasoning seats deliberate only inside what the system computed, and their answers are checked before anything counts.

flowchart LR    subgraph venues["venues, through ccxt"]        direction TB        streams["live streams<br/>trades, order book, liquidations"]        rest["REST history<br/>funding, open interest, mark, index, long/short"]        archives["daily trade archives"]    end    subgraph plane["market data plane: the system"]        direction TB        record["tra record<br/>one bar builder, hour parts, gap marks"]        bars["tra data bars<br/>the same builder over the archive"]        store[("verified archive<br/>Parquet, sha256 per file<br/>flow, book, derivatives, text")]    end    subgraph decide["decision: the system around the deciding team"]        direction TB        judge["Jev, the typed judge<br/>headlines into typed events<br/>reasons checked against the regime"]        features["features<br/>41 figures known at the bar's close"]        readings["readings: macro, on-chain, sentiment<br/>your series against your lines, no model"]        analyst["deciding team<br/>lead, blind confirmer, flow specialist<br/>propose, cite figures"]        gates["gates<br/>cost, exits, paths, confluence, conviction"]    end    shelf[("sources/<br/>dated series you declare")]    subgraph act["risk and execution: the system"]        direction TB        risk["risk gate, signed<br/>caps, loss at the stop, switches,<br/>daily loss, drawdown"]        order["one order<br/>stop and take profit attached"]    end    arena["tra arena<br/>policies replayed, settled net of costs,<br/>beside baselines and a random null"]    streams --> record --> store    archives --> bars --> store    rest --> record    rest --> bars    store --> features    store --> judge --> features    analyst --> judge --> gates    shelf --> readings --> gates    features --> analyst --> gates --> risk --> order --> venues    store --> arena    features --> arena

The shipped cycle reads its own live snapshot; the archive feeds the arena today and the decision path as the work below lands. Where it is going, in the order it is being built:

Each step is measured in the arena before it is trusted, and the four guarantees at the end of this page hold through all of it.

Get started

A few minutes and one model key. Every setting has a working default, so a new desk runs before you change anything; the table under step 6 says what those defaults are.

1. Install. You need Python 3.11 or newer. uv is the quickest way:

uv tool install 'bitra[ccxt,openai,anthropic,google,data]'

That brings every model provider and every exchange in. Without uv, run pip install 'bitra[ccxt,openai,anthropic,google,data]' inside a virtual environment. To move to the newest release later:

uv tool upgrade bitra      # or: pip install -U 'bitra[ccxt,openai,anthropic,google,data]'tra --version

That is the release, and the version to run. The develop branch on GitHub is where the next one is built; a commit there is not a release (more).

2. Make a desk. A desk is a folder holding your settings and everything the agents decide. Name the provider you have a key for:

tra init my-desk --profile autonomous --agent groqcd my-desk

Add --venue okx (or any ccxt id) to trade somewhere other than Bybit.

3. Add your key. Copy the example file and put your key in it. Every command reads the .env next to strategy.toml on its own.

cp .env.example .env        # then set GROQ_API_KEY=... in .env

4. Check. tra check reads every setting against the others and names the one thing to fix next. A desk that is ready ends with:

ok    provider groq: readyready to run `tra run`

5. Run. tra run decides once on live market data, in the sandbox: nothing leaves your machine. tra run --every 15m keeps deciding, once per fifteen-minute bar, on the clock. Or type tra on its own for the console, which takes the same commands.

tra runtra run --every 15m

6. When you want real orders. Start on the exchange's demo or testnet API. Make the desk for that exchange from the start, and the rest is three commands:

tra init my-desk --profile autonomous --agent groq --venue bitgetcd my-desk                       # put the exchange's keys in .env beside your model keytra keys --probe --prepare       # reads the account; on a demo or testnet sets one-way modetra mandate --sign "Your Name" --venue-modetra check

--venue writes what that exchange needs into [execution]: its demo or testnet API, 3x isolated leverage, at most 200 per order, and how its stop and take profit are placed. Change any of it. How to make the key on each exchange is in Exchange accounts and keys, and Toward real orders has the whole path.

What a new desk does before you change anything

Working on the code itself? See Development.

What you see, in order

1. The console. Run tra inside your desk and it opens the console: what is ready, what isn't, and the next command to run. Under it is a prompt that takes every command and draws it right there. Arrow keys go back through what you typed, tab completes, help lists the commands and exit leaves. For scripts and cron, tra -c "run --now" runs one command and exits.

[the console]

2. The pass over your instruments. Type run --every 15m and the desk first refreshes any data series that are due, one line each. Then it looks at every instrument on your list, one data pull apiece, and measures it. The screen steps through them live:

[the screen pass]

3. One cycle per instrument it takes. The system found the candidates, measured them and ordered them by the room their own paths expect over the spread; no model chooses. The instruments taken get a full cycle. The terminal shows the whole pipeline from the start and updates it as each step begins and ends. Steps that run at the same time spin at the same time, each with its own clock, and the line underneath says who is working right now:

[a scheduled desk at work]

4. The result. When the cycle finishes, the table gives way to the pass over the universe: what the system took, what waited and what was held, and why. Then the cycle itself: what the analyst proposed and from which figures, and every refusal with the number behind it. Here an ETH short in a ranging market stopped at its exits: with its own stop and target and the round trip, more than half of trades like it would have had to win just to break even (the break-even bound of the time, an experiment since 0.13):

[one cycle]

Then the data the cycle used and how old it was, the series the readings read, the stop and take profit it would have placed, and which agents answered. A cycle that goes further adds what each reviewer argued, how conviction weighed the voices, the risk at the stop and the targets:

[data, series, exits and agents]

5. Between cycles. The desk says when the next pass is, on the clock and in minutes, for example next pass at 14:30:00 UTC, in 10m 57s, and counts down to it. An auditor woken by an alarm gets a spinner of its own.

The same cycle as plain text (--plain, a pipe or a log): the figures of a real run on live data (2026-09-29), printed by this release with the template's conviction weights, shortened:

cycle 2026-09-29T02-59-16Z BTCUSDT: 5.344s of agents, 15.042s of reviewers  funding        ccxt:bybit         age 4.267s  ohlcv          ccxt:bybit         age 32.746s  orderbook      ccxt:bybit         age 7.544smacro      read by the system: bullish (risk-on)sentiment  read by the system: bullish (positive)lead:      long on range_reversion at 0.65 over 5 bars: the close is stretched low in the windowread: zscore=-0.587, vwap_gap_bps=-3.99, ret_1_bps=1.59reviewer contra    oppose on 2 cited figure(s)reviewer pro       support on 3 cited figure(s)conviction: 0.800 of 1.0 (macro 1.0 of 1.0, sentiment 0.5 of 0.5, structure_lead 1.300 of 2.0)risk: 8.29 bps of equity at the stop against the declared 50 bps, size 0.010cost: 12.87 bps round trip, 2.00 of it estimatedgate: 25.0000 bps claimed clears 12.8723 bps cost plus 0.0000 bps margin  refused targets: the claimed edge is 1.94x the round trip and the desk declared 2.0x; the cost is certain and the edge is not

That trade was refused. It cleared the cost gate and still died, because this desk asks for an edge of twice the cost and got 1.94 times. A refusal names the figure behind it, every time.

How one cycle decides

flowchart TD    find[["the system finds the universe and measures each instrument<br/>one data pull each; every value carries its source and age"]] --> wake    wake{"trigger<br/>has anything moved since the last look?"}    wake -- "no" --> held["held: one data pull, no tokens spent"]    wake -- "yes" --> pick["the system orders what it admitted<br/>the room its own paths expect over the spread"]    pick -- "none taken" --> held    pick -- "taken" --> briefs    briefs["the system writes the briefs<br/>each role sees only the figures its job needs"] --> readers    readers["readings, by the system, no model<br/>macro, on-chain, sentiment against your lines"] --> tech    tech["structure lead, alone or with its team<br/>a blind confirmer that must agree, a flow specialist<br/>a direction, a confidence, a horizon"] --> price["the system prices the round trip<br/>fee, spread, impact, funding, slippage"]    price --> exits{"the stop and the target<br/>the analyst's levels, held inside your bounds<br/>worth more than its cost over its own paths?"}    exits -- "no" --> refused    exits -- "yes" --> round2["reviewers, as many as you declare<br/>a claim that cites no figure is thrown away"]    round2 --> matrix{"confluence<br/>do your declared states agree with the direction?"}    matrix -- "no" --> refused["a refusal that names its figure"]    matrix -- "yes" --> size["conviction<br/>how much, from your ladder"]    size --> risk{"risk per trade<br/>what does it lose at the stop?"}    risk --> gate{"cost gate and targets<br/>does the edge clear the cost, by your margin?"}    gate -- "no" --> refused    gate -- "yes" --> intent["the intent, written into the cycle"]    intent --> venue["venue mode only<br/>one order, the stop and take profit attached"]

Who does what:

The macro, on-chain and sentiment readings are the system's arithmetic over dated series you declare under sources/, and ask no model: until 0.11 each was an analyst whose model labelled the same figures, no decision ever rested on that label, and the calls were retired. Each series is either a file you keep or a feed the desk refreshes itself (FRED, the Crypto Fear & Greed Index, or any paid vendor that answers JSON over https). tra sources show lists every series with where it came from and how old its last value is, and a value older than its max_age is shown as stale and never counted. Data sources has the details.

Which reviewers run is yours to declare. A vote never decides on its own: it moves conviction by the weight you give it ([conviction].reviewer_weight), and a stance counts for what it says beyond its assignment, since a pro that supports is expected to.

Voices are allowed to disagree. Each reading becomes a score between -1 and 1, computed from its own figures against the lines you drew, so a macro reading barely past its line counts as a small voice and one far past it counts as a big one. Conviction adds them up: agreement grows the size, opposition shrinks it in proportion, and only opposition that outweighs the case stops the trade. A short at 0.70 confidence against a barely bullish macro and sentiment scores 0.56 and trades at half size; against a strongly bullish context it scores 0 and doesn't trade.

The structure lead knows what kind of market it is working in before it decides. The system measures the regime on its timeframe (trending up, trending down, ranging or mixed) and on the higher timeframes you list in [run].context_intervals, and it hands over what your macro thresholds say this cycle as well. A setup that doesn't fit the regime, such as a trend trade in a sideways market, is refused by name; [regime] holds the lines.

Everything after the models answer is arithmetic over numbers you declared. No model can place, cancel or resize an order, read a file outside the project, or edit your configuration. Agents has the full table of what an agent can reach, and The cycle walks through every step.

Venues

A cycle reaches venues through ccxt, so every exchange ccxt supports is already integrated: the same order path, the same gates and the same checks, with nothing in the core naming a venue. What differs from one venue to the next is what ccxt reports it can do, the second column, and how far it has been tested, the last:

"Everything" means ccxt reports the five things a safe order needs: send it, attach the exits, read the position, read the open orders, cancel. "Run end to end" is the stronger claim, a full run with real orders on the venue's demo or testnet API, and Bybit, OKX, Bitget, BingX, Binance and Gate have had it so far. How to make a key on each is in Exchange accounts and keys. Any other venue that takes orders can have that run at any time: start on its practice environment, watch a few cycles and read what it sends. Each report moves a venue into the last column.

Adding the venue you trade on is four steps, and one command writes most of it:

tra venues okx            # what it can do, and which keys it signs withtra venues okx --setup    # the blocks to paste into envelope.toml, data.toml, strategy.toml and .env

Your fees are filled in from your account when it reports them, and from the venue's standard tier until then, labelled as such; a fee you write yourself is kept and checked against the account. Venues has the full table, the keys each venue needs, and the four steps.

Models and providers

You choose the provider and the model; nothing is hard-coded. Each provider reads its key from the environment (or the .env beside your config) and nowhere else.

TypeSafe is not in this table because Jev is not a model for [agents]: it is the typed judge, declared in data.toml under [judge] with TYPESAFE_API_KEY (a key from https://console.typesafe.ai/keys) and a pinned version such as jev-1.14.0. tra models typesafe lists the versions your key may use; tra keys --probe proves the key without asking a question.

Whatever the provider, every answer is parsed against a closed schema by the system before anyone reads it, so a model that only offers JSON mode is held to the same shape as one with strict schemas.

[agents]provider = "groq"model = "openai/gpt-oss-120b"effort = "medium"                    # minimal | low | medium | high | xhigh | max, where the                                     # vendor takes it; left out, the vendor's default appliesfallback = ["google:gemini-flash-latest", "anthropic:claude-opus-5-5"]
[agents.roles.reviewer]             # a role can run on its own modelmodel = "a-cheaper-model"effort = "low"

fallback is tried in order when a provider is rate limited or down, with the same brief. A vendor's free tier meters each model on its own, so roles can share a provider across two models and each model keeps its own allowance; tra check judges the load per model and says when one is too much. Every call is a line in usage.jsonl with its tokens and latency, and nothing caps spend unless you declare a ceiling. tra keys --probe proves each key works without sending a prompt. A vendor of your own is about thirty lines (Model providers).

Toward real orders

What it trades today, through one order path and the same gates:

• Linear perpetuals (USDT- or USDC-margined swaps) on any venue ccxt reaches, long and short, with the leverage and margin mode you declare. The default, run end to end on six exchanges.
• Spot pairs, long only and paid in full, for a desk that lists spot in [execution].markets.
• Stock perpetuals and tokenized stocks, which are a perpetual and a spot pair to the venue, with the underlying exchange's hours respected.
• Prediction market outcomes, for a desk that lists prediction, decided on a probability by their own analyst and priced to resolution.

That is where the project started, not where it stops. For spot, Venues says how to set a desk up. Spot, a stock perpetual and a tokenized stock have been run on the Bybit and OKX demos; prediction outcomes have run in sandbox on live data, and their order path is tested against ccxt's shapes until a practice venue is run. Dated futures, inverse contracts and options are on the way, each with its own cost model, exit rules and tests. Until a market type has those, venue mode checks the type on every order and refuses it, because a market the system cannot yet price and protect is not one it should send to. Boundaries says what each one takes.

An agent desk goes through three steps, and you can stay at any of them as long as you like:

Steps 1 and 2 have been run end to end on the demo or testnet APIs of Bybit, OKX, Bitget, BingX, Binance and Gate. Step 3 has not been run by us; treat it as yours to verify, small, first.

If you bring a scripted strategy of your own instead of the agents, it has a research path of its own before any of this: tra data get BTCUSDT 2025-01-01 and tra backtest replay a day of the exchange's archive net of every cost, tra paper --minutes 30 runs it on the live feed with nothing sent, and tra live is a small canary session behind hard caps. The example core is there to show the path, not to make money, and the replay prints its net as it comes out, loss or gain. The live tier has the details.

Venue mode is the only mode that can lose money, so it has a door in front of it. Read the mandate with tra mandate; when you accept it, tra mandate --sign "Your Name" keeps it in the project as MANDATE.md and signs [execution] with your name and today's date, in one step. Signing alone does not switch anything on; --venue-mode signs and turns orders on together. tra check names whatever else is missing.

Your money, your orders, your positions

Once orders flow, three things look after the money, all of them the system's own rules and none of them a model:

Who decides what, in one line each: the analyst proposes the direction, your risk rule sets the size, you choose the leverage, and no model touches the last two. Leverage never makes a trade bigger. It only decides how much margin the trade ties up and how far away liquidation sits.

[execution]mode = "venue"environment = "demo"             # demo | testnet | mainnetleverage = 3                     # yours to choose; it never makes a trade biggermax_notional = 400               # the most one order may be worth, in quote currencymargin_mode = "isolated"         # isolated | crosshold_limit = "timeout"           # close by time after [exit].timeout_bars; or horizon, or none

Underneath, on every venue: the order goes only to the environment you declared (demo, testnet or mainnet); an open position or a working order refuses a second entry; a size off the venue's step is refused rather than rounded; a clock more than a second off the venue's refuses before anything is signed; and Ctrl-C or a service stop cancels a resting order before the process leaves. Every close goes to positions.jsonl and every finished trade to round_trips.jsonl. Capital, orders and positions has every setting, every refusal and the arithmetic behind each one.

One thing to know before you rely on it: the position checks run at the start of each pass of tra run --every. With no desk running, an open position has its venue-side stop and take profit and nothing else, so a desk meant to run unattended belongs under a supervisor.

Orders cross the book by default (taker-taker): they always fill, which is how most desks trade. Point [run].venue_key at a maker-taker channel and entries rest on the book as post-only orders instead - a lower fee, fewer fills, since a resting order can be left behind by the market. Either way the cost gate prices the fee of the channel you chose.

Watching a desk, and what is kept

The console runs commands in its own process, so closing it ends what it was running. A desk that should keep running belongs under a supervisor (systemd, a service manager, tmux), and then you look at it from anywhere with:

tra watch --every 5s       # a read-only view built from the records the desk writes

tra watch opens no connection, holds no key and takes no lock, so opening and closing it does nothing to the desk.

What is kept, and where:

Command line

Every command says what it did. Exit code 0 means cleared, 1 means a gate said no, 2 is a usage error. --json gives machine-readable output and --plain gives text instead of panels.

Getting started

Running

Data

The ladder, research and hosts

What it guarantees, and what is not built yet

Some rules are guarantees, enforced by the system and tests, and every change has to keep them:

• No model touches an order or a key. Models read figures and answer; the system sizes, prices, limits and sends.
• No model can loosen a control. A model can tighten a limit, step the ladder down or arm the kill switch; never widen, promote or disarm.
• One decision, at most one order. A venue error is recorded and the next cycle decides again; nothing is retried around real capital.
• Stale or missing data is refused, never read as zero or as current.

Other things are simply not built yet, and contributions are welcome: inverse, dated futures and options; on-chain venues outside ccxt; more than one account; transfers and treasury; routing and execution algorithms; portfolio accounting. Boundaries lists each one with what it does today and what a contribution would need. The example strategy cores are examples: your edge is yours to declare.

Liability says plainly where responsibility sits when real capital is involved.

Where it is going

The project is in active development, and what this page describes is where it stands today, not where it stops.

• More market types. Spot, stock perpetuals, tokenized stocks and prediction markets are built; inverse and dated futures and options follow, each added with its own cost model, exit rules, sizing and tests before venue mode accepts it.
• More venues run end to end. Every venue ccxt supports already goes through the same order path and gates. Each one joins the six already in the last column of the venues table after a full run on its demo or testnet API, ours or yours.
• The rest of the not-built-yet list, one piece at a time, without weakening a guarantee.

Use the release. What PyPI installs is the latest release, and it is the version to run. Releases come from master. The develop branch is where the next release is built: it moves often, a commit there is not a release, and it can carry work that is unfinished or changed again before it ships. A set of changes becomes a release when it is merged into master, tagged and published, and CHANGELOG.md says what each one brought.

Documentation

The full documentation is at docs.bijectivelabs.dev. Its source is in docs/, so a change in behaviour and the page that describes it go in the same pull request.

Development

git clone https://github.com/BijectiveLabs/bitra && cd bitrauv sync --all-extrasuv run ruff check . && uv run ruff format --check .uv run mypyuv run pytest

CONTRIBUTING.md has the branch flow and the standards the code is held to.

Where it comes from

This repository opened to the public in October 2026. Its history starts on 12 September 2026, when the code was separated from Bijective's internal work, and every commit, tag and contributor since then is here. The ideas it is built on are older: the falsification chain, the cost model and the controls in front of every order come from Bijective's research and from the trading platform it was building before this repository existed. The language-model agents came later and sit on top of those gates, which is why they can propose but never size, price or send.

License

MIT. See LICENSE; NOTICE lists the third-party software this package builds on.

Built by Bijective Labs.