Copilot Project Memory: Keep the Right Work Context

To keep dependable Copilot project memory, maintain an explicit project brief, point the conversation at current sources, and save a checked handoff before you stop. Personal preferences can help shape an answer, but the project still needs its own goal, approved decisions, and next action.

fenbs gives that work a shared home for people and supported assistants. It connects to GitHub Copilot, but not Microsoft 365 Copilot, and it does not import Copilot chats. You deliberately save the reviewed project facts you want the next session to use.

What should you do before the next Copilot session?

1. Name the product and account you will use. A Microsoft 365 conversation and a GitHub Copilot session need different sources and setup.
2. Write a short project brief: goal, intended result, current approved constraints, and what is outside the task.
3. Identify the current source material. Remove ambiguity between accepted documents, old drafts, and examples.
4. Start with a specific task and ask the assistant to explain the relevant constraints before it works.
5. Finish by saving what changed, what was checked, what remains uncertain, and the next useful action.

Use a project already in progress so you can judge whether the record helps. For example, a designer preparing a service page might need the approved audience, a link to the accepted layout, and the fact that the contact form still needs a delivery check. They do not need every abandoned headline from earlier conversations.

The broader Microsoft Copilot memory guide covers personal recall across the product family. This workflow is narrower: arranging the project information you can deliberately check and carry forward.

Where should the project brief and files live?

Keep source documents in the approved place your team already maintains. Write one concise brief that points to the current versions and explains their status. Duplicating a full specification into several memory files creates several places that can become wrong after one revision.

For Microsoft 365 work, Copilot Notebooks↗ (opens in a new tab) let you collect references around a subject. Microsoft says notebook answers use added references that you have permission to access. Check availability in your account and add the approved sources for this project, rather than assuming everything in your organization is included.

You can also provide notebook instructions↗ (opens in a new tab) about focus, format, or tone. Those directions should clarify how to use the source set. They do not establish that an outdated estimate has become the accepted estimate.

For a comparison of contained project spaces, see Copilot Notebooks versus ChatGPT Projects. Keep the immediate setup small enough that you can explain which source governs the task and why.

What belongs in a project brief rather than personal memory?

Personal preferences describe how you generally like assistance. The project brief describes the work everyone needs to understand. A preference such as “explain unfamiliar terms” may apply broadly. A requirement such as “use the approved three-field contact form” applies to a particular deliverable.

Microsoft documents custom instructions for personalizing responses↗ (opens in a new tab) separately from notebook instructions. Put project-specific constraints close to the project’s sources so they do not look like permanent preferences for unrelated work.

• Goal: what outcome the project is meant to achieve and for whom.
• Current direction: the accepted approach, with the person who approved it and its reason.
• Constraints: requirements that change what the assistant should produce or ask about.
• Status: what is complete, what has been checked, and what still needs attention.
• Next action: a small starting step that another person could understand without the previous chat.

In fenbs, the description, tasks, and Decisions and rules keep those distinctions visible; the project context workflow gives a connected assistant a common starting point.

How should coding projects handle lasting instructions?

For GitHub Copilot, keep durable repository guidance in the supported instruction format. GitHub’s repository instructions documentation↗ (opens in a new tab) describes repository-wide guidance in .github/copilot-instructions.md, with other instruction types supported in specific contexts. Check the instructions for the Copilot surface you use.

Good lasting guidance might describe the project’s structure and how to validate changes. Today’s failing check and next debugging step belong in current task progress. Separating them prevents a temporary problem from becoming a permanent instruction that keeps confusing later work.

You can use the same distinction outside coding: a stable editorial rule belongs with maintained guidance, while “the introduction still needs review” belongs with the current task. Our Copilot project management guide covers the wider planning workflow without making personal recall responsible for the project.

How can you check whether the context is working?

Start a fresh session and ask for three things: the current goal, the constraint that most affects this task, and the next unfinished action. Compare the response with the record. If it chooses an old draft, name the accepted source explicitly and correct your project summary if its status was unclear.

Next, ask the assistant to distinguish evidence from assumptions. “The form was built” should not become “messages were delivered successfully” unless that check happened. Save the precise result so another assistant cannot accidentally treat an intended test as a passed one.

When comparing setups across apps, Claude project memory has its own behavior and controls. The portable part of this workflow is the human-readable brief and handoff, not an assumption that product memory settings are equivalent.

How to keep project memory with fenbs

1. Create a free account, free for one person with no credit card, and add the project with a clear goal.
2. Add the current tasks and record the approved constraints in Decisions and rules. Keep proposals open until a person chooses.
3. Connect a supported assistant through Integrations, using MCP, a standard way for an AI app to use an outside tool, and browser sign-in.
4. Ask the assistant to read the project context before starting. At the end, ask it to save task progress and Where we left off, then review the result.

fenbs is not the right fit if you only need a notebook of Microsoft 365 reference files. It does not connect to Microsoft 365 Copilot or synchronize those files. Use it when the project decisions and handoffs need to stay available across supported assistants; the AI memory library explains the surrounding approaches.