Understanding Memory
Factory’s agentic knowledge layer that remembers your workflow and team conventions
Overview
Memory is Factory’s agentic knowledge layer that continuously builds a persistent understanding of your development patterns.
As you work, the assistant quietly records stable facts—branch names, acronyms, release cadences—and places them in either User Memory (private) or Organization Memory (shared). Relevant lines are injected at the start of each chat so replies arrive with built-in context and fewer follow-ups.
Key Benefits
Memory helps Factory understand your context and work more effectively across sessions:
- Work more autonomously – Factory learns your patterns and conventions, so it can complete larger tasks without stopping to ask for clarification.
- Scale tribal knowledge instantly – Solutions discovered by one engineer automatically become available to the whole team. New engineers benefit from accumulated knowledge immediately.
- Focus on solving problems, not explaining context – No need to repeatedly mention that your API uses snake_case or that incidents go to #pagerduty. Factory remembers.
- Context-switch without losing momentum – Your preferences, branch-naming patterns, and project vocabulary persist across sessions and devices.
What Memory May Store
The assistant focuses on facts that are hard to derive from code and that stay useful over time:
- Personal preferences such as explanation depth, coding style, or favourite tools
- Frequently used commands and repeatable workflows
- Team or organisation terms, acronyms, and domain language
- Standard processes, conventions, and best-practice tips
- Pointers to key resources or escalation contacts
- Stable project facts like default branches or staging URLs
Note: Secrets, passwords, and API keys are not saved. A security layer filters them out automatically.
How Memory Works
1. Detect
The assistant recognises statements that will remain useful across sessions.
2. Store
A concise note is saved to User or Organization Memory—no pop-ups, no interruptions.
3. Retrieve
At the start of each session, relevant memory lines blend with live context to guide replies.
| Scope | Visible To | Examples | Who Can Edit |
|---|---|---|---|
| User Memory | Only you | “Prefers bullet summaries”; “default branch = dev” | You |
| Organization Memory | Everyone in organization | “Incident channel = #pagerduty” | Organization Admins |
Working With Memory
Adding New Facts
- Tell Factory directly – e.g. “Remember that our staging URL is <staging_url>.”
- Specify the scope – Optionally, add “for the team” or “just for me” to guide where it’s stored.
Viewing & Editing
- Open the Context Panel and select a Memory context item.
- View the memory content as rendered markdown.
- Switch to code view to edit memory directly. Click the save icon when done. (Editing Organization Memory requires Admin privileges)
Opting Out
- Use the Toolkit Selector to toggle Memory off.
- While disabled, nothing is stored or surfaced for that session; toggle back on at any time.
Downloading a Copy
- Click Download in the Memory tab to save a Markdown snapshot of the current User or Organization Memory.
Security & Privacy
- Encrypted at rest and in transit
- Secret-scrub removes API keys and similar credentials before storage
- Scoped visibility enforces private vs shared access with role-based editing
- Users can opt out at any time via the Memory toolkit toggle
- Lines can be edited or deleted whenever needed