ChatGPT Memory Prompt: A Reusable Context Template

A ChatGPT memory prompt is a block of text you paste at the start of a new conversation to give the assistant the context it needs: who you are, what project you are working on, and any rules that apply. It does not create permanent memory—when the conversation ends, that context is gone—but it gives the session a useful starting point. Tools like fenbs store and maintain that context outside of any one chat app, so you can read the current version into any assistant without retyping it.

The practical value depends on what you include. A prompt with vague preferences ("be concise and friendly") does little. One that names the current project, the active decisions, and the format rules that apply to today's task saves real time at the start of every session.

What a memory prompt does and what it cannot do

Pasting a context block at the start of a conversation puts that information in the model's context window—the text it can see and reason about during that session. It works. The model will refer to what you paste and use it in its answers.

What it cannot do is persist across conversations. Each new chat starts fresh. The built-in memory feature in ChatGPT's paid tiers stores some preferences automatically, but it covers personal preferences and broad habits, not the specific decisions, rules and current status of an active project. For those, a maintained context prompt is still the practical approach. See how ChatGPT project memory works for a comparison of the available options.

Three kinds of information a context prompt carries

Separate the three kinds of information that belong in a context prompt. Mixing them makes the prompt hard to maintain and easy to let go out of date.

• Standing rules: instructions that apply across all work of this type. "All client-facing copy is reviewed by a person before it is sent." "Never include numbers without citing the source." These change rarely.
• Project facts: the current goal, what has been decided, what has been tried. "We are redesigning the onboarding flow. The shorter form was approved on October 10. The confirmation email is still in draft." These change with the work.
• Session task: what you want from this specific conversation. This does not belong in a saved template—write it fresh each time.

Keeping these separate means you update project facts when the work moves forward, without touching the standing rules. The session task stays outside the template entirely.

A template you can copy and adjust

This template covers the two durable sections. Replace the example text with your own information. Keep each section short: two to four sentences is usually enough.

## Standing rules
- All client-facing output is reviewed by a person before it goes out.
- Write in plain prose, no Markdown formatting unless I ask for a list.
- Cite sources when you state a specific number or claim.

## Project: [project name]
Goal: [one sentence describing what this project is trying to achieve]
Current status: [what has been decided, what is in progress, what is waiting]
Active decisions: [list key decisions that apply to today\'s work]
Next step: [what this session is working on]

Notes on each section

Standing rules work best when they are specific enough to check. "Be professional" is vague. "Do not use exclamation marks in client emails" is a rule the assistant can follow and you can verify.

The project status section is where most prompts go stale. After a week of work, the "current status" still says where the project was when you first wrote the prompt. Build a habit of updating this section before you close a session: note what changed, what was decided and what the next step is.

The active decisions list is the most useful part for longer projects. It captures the choices that affect every piece of work in this project: which audience, which format, which constraints. An assistant that reads this list can avoid suggestions that were already considered and rejected.

Where a prompt falls short

A text file or a note you paste from memory tends to drift. You update it when you remember to, not when the project actually changes. Decisions made in one conversation do not automatically appear in the saved version. After a few weeks, the prompt describes the project as it was, not as it is.

A second problem is portability. If you write a context prompt for ChatGPT and then move to Claude or Cursor for part of the work, the prompt does not travel with you. You either re-paste it in every tool or start each session from scratch.

Keeping the template up to date

Whatever format you use to store context, the most important habit is updating it before you stop. At the end of a session, write two or three sentences: what changed, what was decided, and where the next session should start. A prompt that takes two minutes to update at the end saves five minutes of re-explanation at the start of the next one.

Treat the active decisions list as a living record. When a decision is reversed or superseded, mark it as such rather than deleting it. An assistant that sees "we decided to remove the short form (October 10, reversed October 18)" can work with that. One that sees only the most recent change may not understand why it was changed.

How to do this with fenbs

1. Create a free account. Add your first project with a short goal and the current status. This becomes the AI context that a connected assistant reads.
2. Move your standing rules into Decisions and rules. Each rule gets a name, the person who decided it, and the reason. Rules appear first in the AI context every assistant receives.
3. At the end of each session, ask your assistant to save progress: what was completed, what is in progress, and what the next step is. This updates Where we left off.
4. Connect your AI app via Integrations. At the start of the next session, ask the assistant to read the project context before starting work.
5. Add a saved prompt in Skills and prompts for the context block format you want the assistant to use when reading back the project status.

See ChatGPT prompts for work for prompt structures suited to common work tasks, and system prompt examples for templates that hold standing instructions across a session.