Decision record system
Cognivault keeps your team's decisions, the sources behind them, and what actually happened — so six months from now you can prove why, not guess. Every AI answer quotes your own material word-for-word, or tells you it can't answer.
Free plan, no time limit. No credit card required.
Decision Studio — example layout
Decision brief
| Criterion | Weight | Option A | Option B | Option C |
|---|---|---|---|---|
| Revenue impact | ×5 | 4 | 3 | 2 |
| Effort to ship | ×4 | 2 | 4 | 5 |
| Reversibility | ×3 | 3 | 4 | 4 |
| Evidence strength | ×4 | 4 | 2 | 3 |
| Weighted total | 53 | 51 | 54 |
Sensitivity check
If Effort to ship lost most of its weight, Option C would lead instead. Directional, not predictive.
Every quote is matched character-for-character against the source you added, server-side, before it is shown. Anything the model invents is dropped.
What holds this up
Each excerpt is checked word-for-word against your record on the server — twice, once before the answer is built and once before it is stored. Invented quotes never reach the page.
With no matching evidence, or no model available, you get a clearly labelled deterministic draft that asks for the missing fact. You never get a confident guess dressed up as analysis.
Recording a result requires stating what else could have caused it. That one required field is what turns a decision archive into something you can actually learn from.
Four surfaces, one record
Most tools store what you wrote. Cognivault stores why you decided — the options you weighed, the sources you leaned on, the assumptions you named, and what the world did next.
Lay out the options, weight the criteria that actually matter, and score them side by side. A sensitivity check then tells you which single criterion, if it mattered less, would change the leader — so you know how fragile the call is before you make it.
Ask a question and get an answer built only from records this workspace holds. Every quoted excerpt is matched character-for-character against your own source before display; anything the model invents is dropped, not shown with a caveat.
Decisions, beliefs, preferences and assumptions kept as maintained statements with a confidence level and a review date — so a belief that quietly went stale gets flagged instead of silently steering the next decision.
Record what actually happened against what you expected. Cognivault will not let you close a review without naming what else could have caused the result, because a decision log that never records luck teaches you the wrong lesson.
Read this before you sign up
If any of the following is what you are shopping for, this is not the tool — and we would rather you found that out here than after a month of use.
There are no integrations of any kind. Nothing flows in from another tool.
No file upload, no importer, no crawler. You paste in the material that matters and skip the rest. That constraint is the point: a decision record is small on purpose.
Sign-in is email and password. There is no programmatic access and no uptime guarantee.
Search works on the text you put into Cognivault, by keyword. It is not semantic search and it cannot see anything you have not written down here.
Simple pricing
Priced per person, with a monthly AI allowance that grows with the team. Everything you write is yours to keep on every plan.
$0
forever
For one person recording their own decisions
$15
per user / month · $12 billed annually · 3-seat minimum
For a team deciding together
Custom
quoted per organisation
For organisations with procurement or volume requirements
Questions, answered
Decide once, defend it later
Start with one brief, the sources behind it, and a date to check whether you were right.