How Memory works
Goodweek adds what you enter in your profile to every conversation, so you don't have to repeat who you are or how you like to work. Nothing is inferred behind your back: everything comes from what you type or import yourself.
Goodweek reads context from three places. Each has a different job:
Where | What it's for | Limit |
Personal information (Name, Role, Language, Timezone) | Stable facts about you | Fixed fields |
Instructions for Goodweek | How you want Goodweek to respond | 1,500 characters |
Saved memories | Facts about you and your work | 50 lines, 250 characters each |
The golden rule: Instructions say how to respond; Saved memories say who you are. Keeping the two apart gives the most predictable results.
Coming soon: you'll also be able to ask Goodweek to remember, update or forget something directly in a conversation, without opening Settings.
1. Personal information
Fill these first: they take one minute and apply to every chat.
First Name and Last Name: filled in when your account is created. Keep them correct so Goodweek can address you properly.
Role: pick the closest match to what you actually do day to day. Goodweek uses it to pitch answers at the right level (e.g. for a technician prioritizes step-by-step commands, for a service manager prioritizes a summary).
Timezone: detected automatically. Check it is right, especially if you travel or work for clients in other regions. It affects anything involving dates, schedules and deadlines.
Language: your default language. Goodweek replies in the language you write your message in; this setting is the fallback when that isn't clear.
2. Instructions for Goodweek
Use this field for standing rules about how Goodweek should respond: tone, format, length, structure, and things to always or never do. It is applied to every conversation.
Best practices
Be specific and actionable. "Keep answers under 150 words unless I ask for detail" works; "be helpful" does nothing.
One rule per sentence. Short sentences are followed more reliably than long paragraphs.
Say what to do, not only what to avoid. "Use numbered steps for procedures" beats "don't write walls of text".
Avoid contradictions. "Be very detailed" and "keep it short" in the same field produce inconsistent answers.
Keep it general. Rules for one client or one workflow belong in that Space's instructions, not here (see section 5).
Use "Improve with AI" (the wand icon) to restructure a rough draft into clear instructions, then review the result before saving.
You don't need the full 1,500 characters. 5 to 10 clear rules usually work better than a long essay.
Examples for MSP teams
Avoid | Prefer |
I work in IT support. | (That's a fact about you: put it in Role or Saved memories.) |
Be professional. | Write client-facing emails in a friendly, professional tone, without technical jargon. |
Don't give me long answers. | Lead with the answer in 1–2 sentences, then details only if needed. |
Help me with scripts. | When giving PowerShell or Bash commands, put them in code blocks and explain what each one changes before I run it. |
Be careful with security. | Flag any step that could cause downtime or data loss before listing it. |
A good example, ready to adapt
Lead with the answer, then the steps. Use numbered steps for procedures and code blocks for commands. Flag anything that could cause downtime or data loss. For ticket notes, use: Issue / Cause / Resolution / Next steps. Write client-facing emails in plain language, no jargon, under 150 words. If my request is ambiguous, ask one clarifying question before answering.
3. Saved memories
Use Saved memories for durable facts about you and your work: what you're responsible for, the tools you use, how your team operates. Add, edit or delete lines anytime from the Memory page.
Best practices
One fact per line. Each line is read on its own, so it must make sense without the others.
Write complete, self-contained sentences. "Manages the Microsoft 365 tenant migrations for the SMB segment" is clearer than "M365, SMB".
Prefer durable facts. Things that will still be true next month. Skip one-off tasks ("fixing a printer at client X today").
Keep it current. Review your list every few months and delete what's no longer true. Outdated lines lead to outdated answers.
Don't duplicate. If a line says the same as another, or as your Instructions, remove it.
Use the 50 lines wisely. Ten precise memories are more useful than fifty vague ones.
Good examples
Works as a Level 2 technician on the service desk, focused on Microsoft 365 and endpoint issues.
Our PSA is ConnectWise and our RMM is NinjaOne.
Team works on a 4-hour first-response SLA for priority 1 tickets.
Escalations go to the NOC team before reaching the account manager.
Prefers PowerShell over GUI steps when both are possible.
Coming soon: the Saved memories limit increases from 50 to 150 lines, automatic categories (Work Context, Personal Context, Top of Mind, History, Other) that you can change anytime, and smarter selection so each conversation only uses the memories relevant to it. Goodweek will also avoid saving duplicates and update an existing line when something changes.
4. Bringing your memory from another AI
If you already use ChatGPT, Gemini or another assistant, you can import what it knows about you in a few minutes.
On the Memory page, go to Bring your memory from another AI and click Copy.
Paste the prompt into the other assistant and copy its full answer.
Paste the answer into AI answer and click Review adding to memories.
Review every line before confirming. Nothing is saved until you do.
What to do during the review
Delete anything personal that isn't useful at work (family, health, private life), and anything outdated.
Delete anything about a specific client that you wouldn't want in every conversation (see section 5).
Move standing rules to Instructions. Imported lines such as "Always use bullet points" are saved as Saved memories. Rules work best in Instructions for Goodweek, so copy them there and delete the imported line.
Coming soon: the review screen will flag sensitive lines automatically.
5. What not to store
Everything in Instructions and Saved memories is sent with every conversation you have in Goodweek. Treat it as context, never as a vault.
Never store:
Passwords, API keys, tokens, license keys or any other credentials.
Client infrastructure details: IP ranges, hostnames, network diagrams, admin accounts, firewall rules.
Personal data about clients' employees or other third parties.
Sensitive personal information about yourself (health, religion, political views, etc.). It's not needed for Goodweek to help you.
Client-specific context goes in Spaces, not in Memory. If you work on several clients, put each client's context (their stack, conventions, contacts) in that client's Space instructions. That way it's only used when you work in that Space, and it doesn't leak into conversations about other clients.
Coming soon: an automatic safety check will block credentials and sensitive personal data before anything is saved. Until then, the guidance above is the safeguard.
6. Priorities, organizations and control
When things conflict, this is what wins:
Space instructions
Instructions for Goodweek
Saved memories
So a Space set up for a specific client can override your personal preferences, and your Instructions override anything in Saved memories.
Memory is separate per organization. If you belong to several Goodweek organizations, each one has its own Instructions, Saved memories and Role. Nothing carries over between them, by design, so one client's context never appears in another's. Set up your Memory in each organization you work in.
Response language. Goodweek answers in the language of your message. You can write Instructions and Saved memories in any language: the language they're written in doesn't change the language Goodweek responds in.
You're in control. The Remember details across chats toggle pauses Memory at any time. Turning it off doesn't delete anything: your Instructions and Saved memories are kept and are used again when you turn it back on. To remove something permanently, delete it from the Memory page.
Coming soon: turn Memory off for a single conversation and an organization-level switch for admins to enable or disable Memory for everyone (admins won't see memory content).