Project knowledge
The durable memory of your project — facts, assumptions, decisions, and metrics that ground every AI answer.
Project knowledge is a persistent store of context that AI skills read before generating anything. It's the difference between generic output and output that sounds like it was written by someone on your team.
What to put in it
Short, factual, current statements:
- Goals — "Increase activation of self-serve signups to 40% by Q4."
- Constraints — "No engineering capacity for native mobile before 2027."
- Decisions — "We chose usage-based pricing (decided 2026-05-14)."
- Key metrics — "Current churn: 3.1% monthly; NPS 42."
- Definitions — "'Activation' means first canvas shared within 7 days."
How it's used
When you run a skill or ask a question, Crux selects the knowledge entries most relevant to your request and includes them in the AI context, alongside canvas and file context. The Help agent also reads your plan and workspace context when you ask product questions in the app.
Where to find it
Open Project knowledge from the sidebar rail. Entries can be added by hand, and some skills propose knowledge updates (for example after a decision is recorded on the canvas) — always subject to your review.
Keep it pruned
Stale knowledge is worse than no knowledge: review the list at milestones and delete or update entries that no longer hold. Aim for a screenful of high-value statements rather than an archive.
Related articles
Documents: Project knowledge
