AI Instructions let you give the Keypup AI Agent persistent knowledge about your team's habits, workflows, and conventions.
Instead of repeating context in every conversation, AI Instructions ensure the AI consistently applies your established patterns when generating insights and configuring widgets.
Why use AI Instructions?
The Keypup AI Agent works best when it understands how your team operates. Without AI Instructions, the agent relies on generic assumptions.
With them, it can:
Interpret your labels and statuses correctly (e.g., knowing that
priority:highmeans a critical item in your workflow)Understand your sprint cadence and delivery expectations
Recognize your branching strategy and code review process
Provide insights that match your team's actual definitions of "done", "in progress", etc.
Think of it as onboarding the AI to your team, once.
How to configure AI Instructions?
AI Instructions are accessed from the Team Management screen
Go to Settings > Team Management
Click the AI Instructions tab
Write your instructions in the markdown editor
Click Save
Only team admins can edit AI Instructions. Non-admin members can view the current configuration but cannot modify it.
The editor supports full markdown formatting (headings, bullet points, bold, etc.), which helps you structure your instructions clearly. There is a 10,000-character limit.
Writing effective AI Instructions
The key is to describe what the AI can't infer on its own. Focus on conventions, definitions, and patterns that are specific to your team.
Here are the areas worth covering:
Work patterns
Sprint or iteration length (e.g., 2-week sprints)
Release cadence (e.g., weekly releases on Tuesdays)
Code review requirements (e.g., minimum 1 approval required)
Key labels and statuses
Map your labels to their meaning:
priority:critical= production outage,tech-debt= non-urgent refactoringDefine your workflow statuses:
Backlog→In Progress→In Review→QA→DoneClarify any custom statuses or labels unique to your team
Git branch conventions
Branch naming patterns (e.g. feature/JIRA-123-description, hotfix/description)
Main/default branch name (e.g. main, develop)
Merge strategy (squash, merge commit, rebase)
Team structure
How teams or squads are organized
Which projects or repositories belong to which team
Definitions
What counts as "cycle time" for your team
How do you define "lead time" or "deployment frequency"
Any custom metrics or KPIs your team tracks
Example of AI Instructions
## Work Patterns
- Sprint cycle: 2 weeks (Monday to Friday)
- Code reviews: minimum 1 approval required
- Releases: every Thursday
## Key Labels
- priority:high - Critical, needs immediate attention
- priority:low - Nice to have, backlog
- bug - Production issues
- tech-debt - Refactoring tasks
## Git Branches
- Feature branches: feature/JIRA-XXX-short-description
- Hotfixes: hotfix/short-description
- Default branch: main
## Workflow
Backlog > In Progress > In Review > QA > Done
## Definitions
- Cycle time: from first commit to merge into main
- A PR is considered "stale" after 3 days without activity
Tips
Here are a few recommendations when writing AI Instructions:
Be specific rather than generic. "We use 2-week sprints" is more useful than "We follow agile."
Use markdown headings to organize sections. The AI parses structured content well.
Update your instructions when your process changes. Outdated instructions can lead to misleading insights.
You don't need to describe what Keypup already knows (field names, chart types, etc.). Focus on what's unique to your team.
The AI Instructions are shared across your entire team, so every team member benefits from the same context when using the AI Agent or AI Assistant.