Your trade journal writes itself from your wallet history.
Every closed position becomes an entry with a reconstructed chart, a letter grade, tagged mistakes and one paragraph of analysis, without you typing anything.
You have kept a trading journal before. A spreadsheet, a Notion table, maybe a paid one with a CSV importer that wanted a clean export from three different bots. It died on the first busy day, when the trades stacked up and you did not feel like typing them out. Everything you added after that was reconstructed from memory, which is to say it was fiction.
This journal reads the chain instead of asking you. You connect a wallet with one off-chain message signature, never a transaction, or start with email and attach a wallet later. From there Dossier reads every closed swap that wallet has made inside your plan's history window. New trades land in seconds. The graded post-mortem usually follows in under a minute, and close-to-grade is under 45 seconds at p95.
The practical difference is coverage. A hand-written journal contains the trades you were willing to write down, which are rarely the ones costing you money. This one contains all of them, including the 3am revenge buy you would prefer not to look at again.
Positions, not swaps
A raw swap log is not a journal. You scale into a token three times over four minutes, trim a third into the first pump, trim again, then dump what is left an hour later. That is six on-chain transactions and one decision.
Dossier groups buys, sells and partial exits into a single closed position. Those six transactions become one journal entry, one grade, one paragraph. The entry is written when the position closes, not when a swap lands. Positions that are still open sit in the Live tab until they go flat.
What one entry contains
Every entry opens on a reconstruction of the trade, not on a screenshot you forgot to take.
The coach reads the chart, the token metadata, entry timing, exit timing, your recent trades and a rolling 90-day profile of how you trade. The figures on the entry are computed from your data. The model interprets them. It does not produce them.
- The chart: 1-minute candles, 15-second bars on scalps, from 15 minutes before your entry to 60 minutes after your exit, on an interactive chart.
- Entry and exit timing: where your buy and your sell actually sit on that chart.
- Token provenance: bundler-launch flag, launch time, deployer wallet, and a 1-10 risk score.
- The grade: one letter, A to F.
- Tagged mistakes: the codes this trade matches, from the fixed list of ten.
- The coach paragraph: one paragraph, no fluff, no emoji.
The grade is about the decision
A to F, and nothing else. Not a 0-100 score, not a trade quality index, no radar chart with five axes you would never look at twice. A letter sorts, and it is hard to argue with.
Grades judge decisions, not outcomes. A trade that printed can grade badly, because you sized it at random and got paid anyway. A losing trade with a planned entry, a stop you respected and an exit at your level can grade well.
Exits are judged only on data that was visible before the sell. If the token ran again long after you were out, the grade does not know and does not care. A coach with hindsight is useless, because from the future every exit looks stupid.
Repeat mistakes carry a count
The taxonomy is fixed at ten codes: FOMO_ENTRY, CHASED_PUMP, REVENGE_TRADE, NO_STOP, MOVED_STOP, BAG_HELD, PANIC_SOLD, EXIT_TOO_EARLY, ENTRY_TOO_LATE, SIZE_TOO_BIG. Fixed matters. Free-text tags turn into a pile of different descriptions of the same problem, and nothing you can count.
From the second occurrence, a mistake is flagged as a repeat and carries a running count. One FOMO_ENTRY is a bad night. FOMO_ENTRY as your most common tag is not a bad night, it is a description of how you trade, and you cannot see it from inside a single post-mortem.
The Trader Report Card
Single entries tell you about single trades. The Report Card is the aggregate: one page, recomputed on every graded trade. It names your worst hour, your worst category, and your single biggest leak. One leak, not a long list of things to work on, because a long list gets ignored the same way the spreadsheet did.
What it surfaces is usually boring and specific. An hour of the day. A category of token. A size you only reach for when you are already down. Boring and specific is the part you can act on.
Export and deletion
Export the journal to CSV for a spreadsheet, or to Obsidian-ready Markdown if that is where your notes live. It is your trade history and it should not be stuck here.
Deleting a wallet hard-deletes its trades, positions and analyses immediately. Not a flag on a row, not a 30-day grace period.
How far back the journal reaches depends on the plan: 30 days and 15 analyses a month on Free, 90 days and 150 on Pro, 365 days and 500 on Degen.
What it does not do
- No CSV importer. There is no import step to fail and no bot export that has to match a template. Data only moves outward.
- No buy or sell signals, on a journal entry or anywhere else. Every position in the journal is already closed.
- No 0-100 score, no trade quality index, no radar chart. One letter, A to F, and the reasons behind it.
- No hindsight in the exit grade. What the token did after your sell is never used against you.
- Solana mainnet only. No other chains, no centralised exchange fills, no mobile app and no browser extension.