What’s actually slowing this PC down?
Pick the symptom - the matching free tool is one click away.
You can give ChatGPT and Claude reusable context for ongoing work by keeping three small files: profile.md for stable background, project.md for current work, and instructions.md for how you want responses delivered. Add the files to a project in each service, then put the most important behavior rules in that service’s project-instructions field too. This is a practical organization method, not a vendor-prescribed template.
What the three files should contain
Keep durable information separate from details that change often. That makes it easier to update a project without accidentally replacing your general preferences or response rules.
As an Amazon Associate I earn from qualifying purchases.
profile.md: stable background and preferences
Include information that helps the assistant tailor work across tasks, such as your role, audience, recurring goals, preferred terminology, and constraints. Avoid sensitive personal details unless they are necessary and you have checked the service’s data and workspace settings.
# Profile
- Role: [your role]
- Audience: [who the work is for]
- Goals: [recurring goals]
- Preferences: [stable preferences]
- Constraints: [important boundaries]project.md: facts and decisions for the current project
Describe the project’s purpose, audience, key facts, terminology, decisions, and links to authoritative source material. Mark facts that may go stale with a date, and distinguish confirmed decisions from open questions.
#1 Best Overall
# Project
- Purpose: [what this project is meant to achieve]
- Audience: [who it serves]
- Current facts: [verified details]
- Terminology: [terms and definitions]
- Decisions: [decision — date]
- Open questions: [unresolved items]
- Sources: [authoritative links]instructions.md: response style and workflow
Write down how the assistant should approach the work: tone, level of detail, preferred structure, how to handle uncertainty, and any required output format. Keep these instructions concise. Put the highest-priority rules in the project’s built-in instruction field as well, so the assistant has clear guidance without relying only on a file being consulted.
# Instructions
- Write for: [audience]
- Tone and detail: [style]
- Workflow: [steps to follow]
- When uncertain: [how to qualify or ask]
- Output format: [required structure]Replace the bracketed examples with your own content; they are prompts for what to write, not required syntax. Do not put passwords, API keys, or other secrets in context files.
Set up persistent context in ChatGPT
- Create or open a project: In ChatGPT, use Projects to group the chats and reference materials for this work. The project keeps its files and instructions together. See OpenAI’s Projects guidance for current interface details.
- Add the files: Upload the three context files to the project’s files or reference-material area. OpenAI’s documentation lists TXT among supported text formats and describes adding reference files or pasted text. It does not specifically guarantee that every interface accepts the
.mdextension, so if an upload is rejected, use a.txtcopy or paste the contents. See OpenAI’s file-upload guidance. - Add project instructions: Open the project’s instructions control and paste the essential rules from
instructions.md. Keep detailed reference material in the files; use the native field for the few rules that most affect how the assistant responds. - Choose the memory behavior deliberately: ChatGPT Projects may use default memory or project-only memory, subject to personal and workspace settings. In project-only mode, ChatGPT does not refer to saved memories or conversations outside that project, although chats within the project may refer to one another. Check the project’s available memory setting before relying on a particular boundary.
Set up persistent context in Claude
- Create or open a project: Use Claude’s Projects feature for the work. Anthropic currently describes Projects as available to paid-plan users; eligibility can change. See Anthropic’s project-management guidance.
- Add the files to project knowledge: Upload the context files to the project knowledge area. Anthropic says information uploaded there is used across chats in that project. Its upload documentation lists TXT among supported document types but does not specifically guarantee a
.mdextension in every interface; if needed, try a.txtcopy or paste the text. See Anthropic’s document-upload guidance. - Add project instructions: Put the essential response and workflow rules in the project’s instructions field. Project knowledge supplies reusable reference information; instructions guide how Claude should use it.
- Keep cross-chat context in knowledge: Claude does not automatically share one chat’s context with other chats in the project. Add decisions or facts that should carry across chats to project knowledge, such as by updating
project.md. Anthropic says retrieval-augmented generation (RAG) activates automatically as project knowledge approaches the context-window limit and can expand capacity by up to 10x; that is Anthropic’s product statement, not an independent performance measurement. See Anthropic’s explanation of Projects.
How context and limits differ
Both products combine project instructions with reusable material, but their cross-chat behavior is not the same. ChatGPT’s project memory settings can let chats refer to other chats in the project. For Claude, put information in project knowledge if you want it available across chats.
| Detail | ChatGPT Projects | Claude Projects |
|---|---|---|
| Cross-chat context | Project chats may refer to other chats in the project, depending on memory settings. | Information must be added to project knowledge to be shared across chats. |
| File count or size | OpenAI’s current Projects guidance lists 5 files per project for Free; 25 for Go and Plus; and 40 for Edu, Pro, Business, and Enterprise. Up to 10 files can be uploaded at one time. Confirm current entitlements in OpenAI’s Projects guidance. | Anthropic lists a 30 MB maximum per file and no fixed project-knowledge file-count limit, subject to the context window. RAG may expand capacity by up to 10x as that limit approaches. Confirm current details in Anthropic’s upload guidance and Projects guidance. |
| Plan eligibility | File allowances vary by plan as listed above. | Anthropic currently describes Projects as available to paid-plan users. |
These are vendor specifications, not independent benchmarks. Plans, supported extensions, interface labels, and limits can change, so check the linked help pages and your account before building a workflow around a particular allowance.
Rank #3
Check sharing and data-use settings before adding sensitive context
Project files make information easier to reuse, but they are not by themselves a confidentiality guarantee. Review the account and workspace controls that apply to the people who can access the project.
Quick Recap
- ChatGPT: Shared projects automatically use project-only memory and cannot access a member’s outside context, custom instructions, or memories. OpenAI says Business, Enterprise, Edu, and Healthcare workspace data is not used to train models by default. For consumer Free, Plus, and Pro plans, project information may be used if “Improve the model for everyone” is enabled. These settings and plan distinctions are described in OpenAI’s Projects guidance and file-upload guidance.
- Claude: Anthropic’s project documentation describes private or organization-wide visibility for Claude for Work, governed by available workspace settings. Check those controls before adding personal or company information. See Anthropic’s project-management guidance.
Maintain the files as the work changes
- Update the project file when facts or decisions change. Record the decision and date, and remove superseded details so old guidance does not compete with the current version.
- Date volatile facts. Add an “updated” date to information such as schedules, policies, or product details that may change, and verify it before a later task depends on it.
- Keep durable and temporary context separate. Move a preference into
profile.mdonly if it should apply broadly; keep project-specific facts inproject.md. - Trim instructions to the rules that matter. If a rule is essential, put it in the native project-instructions field; keep examples and detailed reference material in the files.
- Review project access after sharing changes. Remove material that collaborators no longer need and revisit workspace controls when membership or account settings change.
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
