Guide
How to give Claude, ChatGPT and Copilot your company’s context
Every AI assistant starts a conversation knowing nothing about your company. There are four ways to fix that, and they differ in who does the upkeep, who can see what, and whether the context is still true next month.
In short
You can paste context into each prompt, keep a context file such as CLAUDE.md or custom instructions, rely on built-in memory, or connect a shared context source over MCP. The first three depend on upkeep by hand and stay with one person or tool. For a team, MCP serves the same current, permission-scoped context to Claude, ChatGPT, Copilot and Cursor.
Updated
Four ways, compared
The methods differ less in set-up effort than in upkeep. Whichever you choose, someone or something has to keep the context true as decisions change.
- Pasting into the prompt: quick for one question, gone when the chat ends, and only as good as what you remembered to paste
- Context files and custom instructions (CLAUDE.md, AGENTS.md, a ChatGPT Project): reusable, but a snapshot that drifts within weeks and carries no permissions
- Built-in memory: remembers one person’s preferences inside one vendor’s tool, with no sources you can audit
- A shared context source over MCP: one record every MCP-capable tool can read, kept current at the source rather than by you
Why MCP is the team answer
The Model Context Protocol is an open standard for connecting AI tools to data. Connect a context source once and Claude, ChatGPT, Copilot and Cursor read from the same place, so answers stop depending on who wrote the prompt. Copilot support depends on the MCP capabilities of your deployment, so check that before you plan around it.
What the context source needs
MCP is only the connection. What matters is the record behind it.
- Captured where decisions happen, in meetings and threads, rather than typed in afterwards
- Stamped with when each fact was true, so the AI can say “as of Tuesday’s review”
- Scoped by permission, so each person’s AI sees only what that person may see
- Traceable to its source, so a doubtful answer can be checked against the meeting it came from
How In Parallel does it
In Parallel captures decisions, risks and commitments from your meetings and threads into workspaces and serves them over MCP. Reads from your own AI tools are free up to 10,000 per user a month, pooled across the company, and AI agents need no seat. Keep a context file for tool behaviour and repo conventions if it helps; the company context comes from the record.
FAQ
Common questions
- Can I connect ChatGPT or Claude to my company’s knowledge?
- Yes, through an MCP server, which Claude, ChatGPT, Copilot, Cursor and other MCP-capable tools can read from. In Parallel publishes your company’s decisions, risks and commitments as an MCP server, scoped to the workspaces each person may see.
- Is a CLAUDE.md file enough context for a team?
- For one person in one repo it can be. For a team it drifts within weeks, gives everyone the same ungoverned context and cites no sources. Keep it for tool behaviour and conventions, and take company context from a source that stays current.
- Which AI tools can read In Parallel’s context?
- Any MCP-capable client, including Claude, ChatGPT, Copilot and Cursor. Copilot support depends on the MCP capabilities of your deployment. Reads are free up to 10,000 per user a month, pooled across the company.
Start with your next meeting.
Bring your team in. Use every surface, including every other team's. From €29 per user per month, billed monthly, no commitment.
30-minute demo. We'll show you how meeting signals become a shared record of decisions, owners and risks.