Running it without you

In short: you want something to happen while you are away, like buying when the price drops to your number, setting a stop loss, or collecting rent every month. In chat, the agent writes this as a strategy card: English rules that wait for a condition, a price or a clock. A workflow can also start on a schedule, at a price you set, or when the workflow before it finishes. Each time it starts, the plan is built again and priced at that moment. Switching one on is your own decision, because from then on it spends real money without asking.

​

Cards and workflows

A strategy card is the main way to automate something in chat. Describe the rule and the agent writes the card, backtests it and shows you what it can do. You approve it with your signature, start it, and stop it when you want. See Strategy cards, what a card can watch and do and going live. The chat writes cards only. It does not write workflows, schedules or chains. The three mechanisms below still exist and still run. You can set them up from the console, in the plan designer and on the swap page, and outside agents set them up through the tool server. A card carries its own limits and its own budget, and a new version needs a new approval. A workflow rebuilds and re-approves its plan at every firing instead. The rest of this page describes workflows, schedules, price triggers and chains.

​

They all work the same way

A schedule and a price trigger both point at a workflow they do not contain. Neither agrees to anything in advance. Each one rebuilds the plan from scratch every time it fires. That is what makes leaving one on for months reasonable. When the clock strikes or the price arrives, the plan is rebuilt, priced against the market at that moment, and approved then. The limits you set apply to that day’s numbers, not to the numbers of the day you set it up.

​

Switching one on is its own decision

A saved workflow does nothing. A saved schedule does nothing until you switch it on. Switching it on is deliberately separate from writing it, because from that moment it fires unattended and spends real money every time. An outside agent can write all three and cannot switch any of them on. Setting a price trigger and switching on a schedule need the same approval from you as pressing Run. A schedule is in one of three states: Off and paused look similar. One waits on software, the other on you.

​

When a firing cannot run

The same workflow runs once at a time. So a monthly firing can arrive while the previous month’s run of it is still going. It can also arrive while that run is stopped at an approval gate, waiting for you, which waits indefinitely. Two different workflows on the same account do not collide. A run that went stuck at nine in the morning does not stop an unrelated strategy at four in the afternoon. The firing is recorded rather than dropped, and it asks you which you meant.

​

A price trigger fills in pieces, and that is normal

A trigger rests as a price and a size. When the price arrives, the plan is rebuilt sized to what is left. A fill of 40% leaves the other 60% resting, and the trigger re-arms and keeps waiting. A trigger can take several shapes: wait at a level, hold in a band without chasing, a stop, a trailing stop, or slice the size out over time. Order types describes each one and when to use it. A large order against a pool that cannot absorb it in one go is supposed to arrive in pieces. Each piece is separately rebuilt, separately approved, and carries its own price floor enforced on the network. A trigger ends by itself when it is fully filled. Until then, “what am I waiting on” is always answerable.

​

A chain is workflows in order

A chain lets you say “do this, then keep doing that” without forcing two different rhythms into one document. Buying a property runs once. Splitting its rent between the debt and more collateral runs every month. You edit, approve and re-run those on completely different schedules, so they are separate workflows with a chain naming the order.