- November 19, 2026
Claude Projects, ChatGPT Projects, or an MCP Connector: Which Keeps Your Messaging Consistent?
Projects give AI your instructions and files. An MCP connector gives it access to a source. How each keeps team messaging consistent, and when to combine.
Claude Projects, ChatGPT Projects, and custom GPTs keep messaging consistent by giving the AI instructions and reference files. An MCP connector doesn't, on its own: it gives the AI access to whatever sits at the other end. So the useful question is what you connect to. A Project or a Skill is enough for one person with stable documents. When a team shares them, context goes stale, one project becomes a junk drawer, and every tool holds its own copy of your positioning. That is where a context layer earns its place: one governed source that your Projects and your other AI tools all read from over MCP.
What is the short answer?
- One person, one tool, stable documents: a Claude Project or a ChatGPT Project is enough.
- A team in one tool: a shared Project, or a shared Claude Skill, works if someone owns it.
- A team across tools, or more than a brand guide: connect a source over MCP, and keep the Project or Skill for the instructions.
- Most teams end up with both: instructions in a Project or Skill, facts in a connected source.
What does a Claude Project do?
A Claude Project is a self-contained workspace with its own chat history and knowledge base. You upload documents, and you write project instructions. Claude uses what is in project knowledge across all of the chats in that project. Context is not shared between chats in a project unless it is added to the knowledge base.
Limits that matter for messaging: project files can be up to 30MB each, the number of files is not capped but the total has to fit in Claude's context window, and Claude extracts text only from most documents. Paid plans scale with retrieval (RAG) when a project's knowledge nears the context limit. On Team and Enterprise plans, a project can be shared with a "Can view" or "Can edit" permission. Anthropic describes projects as providing static background knowledge that is always loaded when you start chats in them.
What does a ChatGPT Project do?
A ChatGPT Project keeps related chats, files, and instructions together, and ChatGPT can use that context to support ongoing work. Project instructions apply only within that project and override your global custom instructions. File limits depend on the plan: 5 files on Free, 25 on Go and Plus, and 40 on Edu, Pro, Business, and Enterprise, uploaded 10 at a time.
Projects can be shared, and OpenAI lists content creation among the uses, applying project-specific instructions to keep tone and style consistent across contributors. OpenAI also describes shared projects as ones that evolve over time due to sustained user contributions. That helps a team work together. For a single source of truth, it means someone still has to decide what stays in the project.
What about custom GPTs?
A custom GPT holds instructions and up to 20 knowledge files of up to 512 MB each, so it can work as a brand assistant. OpenAI is retiring them:
- Custom GPTs are scheduled to retire on December 11, 2026 (February 11, 2027 for FedRAMP workspaces and workspaces with an approved deferral).
- New GPTs can't be created or published on personal Free, Go, Plus, or Pro accounts.
- OpenAI offers migration to plugins. A GPT's instructions become a skill inside the plugin, its knowledge files are copied in as reference files, and custom actions do not transfer.
If you have a brand GPT, treat the migration as the moment to decide where your guidelines should live for good.
What does a Claude Skill do?
A Skill is a folder of instructions, scripts, and resources that Claude loads when it is relevant to the task. Anthropic lists applying brand style guidelines to documents and presentations as a use. Where a Project is static background knowledge that is always loaded, a Skill is a procedure that Claude brings in when it needs it. On Team and Enterprise plans, a Skill can be shared with specific people, and recipients automatically get your updates at next use. A skill you upload stays private until you share it.
What does an MCP connector do?
MCP, the Model Context Protocol, is an open standard for connecting AI applications to tools and data. Anthropic describes remote MCP servers as giving models access to internet-hosted tools and data. In Claude, a connector lets Claude work with the tools and data sources in your workflows. In ChatGPT, apps connect ChatGPT to services so you can work with their information in a conversation, and plugins can bundle apps with skills.
A connector is access, not instruction. It does not tell the AI how to write. What it carries is whatever sits behind it: if that is a stale document, you have connected a stale document. Anthropic puts the split this way: MCP connects Claude to external services and data sources, while skills provide procedural knowledge, and you can use both together.
How do they compare?
| Claude Project | ChatGPT Project | Claude Skill | MCP connector | |
|---|---|---|---|---|
| What it gives the AI | Your instructions and uploaded files | Your instructions and uploaded files | Instructions and resources, loaded when relevant | Access to tools and data hosted elsewhere |
| Where it works | Claude | ChatGPT | Claude | Any AI tool that supports MCP |
| Shared with a team | Team and Enterprise, view or edit | Shared by invitation or link, chat or edit access | Team and Enterprise, view-only | Each person connects; on Claude Team and Enterprise, an admin adds it once first |
| What keeps it current | Someone replaces the files | Someone replaces the files, and members can add to it | The owner updates it, recipients get the update at next use | Whatever you connect to |
Which one keeps messaging consistent for a team?
None of them does it alone. The tools differ in what they hold, and consistency comes from who owns the material.
- Projects and Skills carry your instructions well. Voice, audience, banned words, and examples fit in instructions and a few files. Use them for that.
- They are weaker as the home of your facts. A product fact copied into three Projects is three places to update. The same is true of a custom GPT's knowledge files.
- A connector carries the facts, if you give it a source worth connecting. Point it at a governed source and every tool reads the same answer.
- Combine them. Keep the how in the Project or Skill, and the what in the source.
Do I need a separate connector for each AI tool?
You add the connection once in each AI tool, and they all read from the same source. Marcora's connector, mcp.marcora.ai, works with Claude, ChatGPT, Cursor, and any AI tool that supports MCP connections. Update a fact in Marcora once, and every connected tool reads the change.
Where does Marcora fit?
Marcora is the context layer for go-to-market teams, and it is both the governed content and the connector. Your Brand Foundation (positioning, brand voice, and writing style) is always included. A Reference Library holds product docs, buyer personas, competitive analysis, and messaging, scored for the audience you are writing for. You can add a campaign's material to a Project in Marcora, so it applies there and stays out of everything else.
Marcora is listed in Claude's connector directory. Connect Marcora in ChatGPT from the plugin directory. You'll need a Marcora account. Setup takes about 2 minutes, the setup time listed on the Claude connect page and the ChatGPT connect page. For the steps, see How to Connect Your Brand Guidelines to Claude and ChatGPT.
Frequently asked questions
Does Claude use a Project's files in every chat?
Anthropic says anything you upload to a project's knowledge is used across all of your chats within that project. Context from one chat is not shared with another unless it is added to the knowledge base.
Does ChatGPT always use a connected app?
OpenAI says ChatGPT automatically uses your installed plugins when they are relevant to your request, and that it may ask you to confirm before an app reads information or completes an action. You can call a plugin directly by typing @ and its name.
Can I use a connector inside a ChatGPT Project?
Yes. OpenAI says you can use connected apps in project chats, and ChatGPT may ask you to confirm before an app searches outside the project.
Are custom GPTs going away?
OpenAI says custom GPTs are scheduled to retire on December 11, 2026, with migration to plugins. Existing conversations with custom GPTs remain accessible after retirement.
Can I use a Skill and a connector together?
Yes. Anthropic says you can use both together: MCP connections give Claude access to tools, while skills teach Claude how to use those tools effectively.
Is Marcora a replacement for Claude or ChatGPT?
No. It works with the AI tool you already use. The AI writes; the context layer makes sure it writes from the right material.
See it work: How Marcora assembles context before the first word is written, or connect it in Claude or ChatGPT. Related: What Is a Context Layer for Go-to-Market Teams?.

