Most chart assistants start every conversation as a stranger. Wickline carries four kinds of memory into every message: the rules you wrote, the read you are carrying, the setups you plan, and the record of what you are actually good at. You never manage any of it. You just notice the agent was already there for last week’s session.
They do different jobs on different clocks. A rule outlives everything. A thesis lives until price proves it wrong. A commitment lives until its trigger fires. The record only grows.
Account-level laws that outlive every chat, every symbol, and every model. Risk limits, instrument bans, position sizing. Write one once and it is in the room for every trade after it, so the agent argues from your rules instead of generic risk advice.
One live view per market: a stance, a conviction, the reasoning in a line, the price you wrote it at, and the level that would prove you wrong. The big picture is always there. You can be looking at a bullish 5m and the agent still knows you are in a strong bearish daily downtrend.
A decision with a trigger attached. "No longs until a 4H close above 83." The condition is machine-checkable, not a vibe: a predicate, a level, and a timeframe, so the agent knows precisely when your own rule releases you.
Every logged trade becomes evidence. Sliced by setup, direction, timeframe, and symbol, so the agent can tell you which version of you is profitable before it endorses an entry.
“No longs until SOL closes a 4H above 83.” That isn’t a note in a file. It’s a predicate the agent evaluates against live price.
You dictate it, or the agent proposes one in structured form and you say yes. Never created quietly in the background.
A predicate, a level, and a timeframe: close above 83 on the 4h. The agent knows exactly when your own rule releases you.
The commitment is in the prompt on every message about that market. The ones that stop mattering expire on their own; the ones that still do come back for review.
You can always trade against it. The decision is recorded either way, and the ratio becomes a number you can look at.
One standing read per market: a stance, a conviction, the reasoning in a line, the price you wrote it at, and the level that would prove you wrong.
The moment price runs through the level you said would prove you wrong, the agent says so, before it reasons from the read. You find out your view broke from the agent, not from your P&L.
Every thesis stores the price it was written at, so the agent can say “you wrote this at 3,480 and we are 6.1% away” instead of vaguely gesturing at a chart.
Replacing a thesis does not delete it. The previous stance, conviction and reasoning go into a flip log you can read back. The exact moment you turned from bearish to bullish is in there, with the reason you gave, the time, and the price you did it at.
Your BTC stance and your ETH stance are separate records with separate clocks, and the agent sees the whole board when it reasons about any one of them.
Memory that can hold you to your rules is only trustworthy if it can’t write them. These are constraints in the product, not promises in a prompt.
It can tell you it has watched you break the same rule twice. It cannot turn that into a rule you have to follow. That switch is yours, and it stays off until you flip it.
A commitment exists only if you dictated it, or if the agent proposed one in structured form and you said yes. There is no ambient promise-making.
You can always trade against your own rule. The agent won’t stop you. It records the decision so the honored/overridden ratio stays honest.
No exchange keys, no order routing, no custody. Memory shapes what the agent says. Execution stays entirely with you.
See what it can tell you about yourself that your spreadsheet never did.