Connect Creed to Cursor
To connect Creed to Cursor, add it as a remote MCP server in Cursor's settings and authorize it once with OAuth. Point the server at https://creed.md/mcp, then complete authorization in the browser by clicking Allow on the Creed consent screen while signed in to creed.md. There is no key to copy; the handshake is OAuth.
Once connected, Cursor can read your Creed profile before it plans or edits, so it starts already knowing your role, stack, goals, and preferences rather than working cold. As it learns something durable about you, it can propose a narrow update you approve.
Creed is one personal context file that every AI reads before it answers. In Cursor, that means the same profile your other agents use also shapes how Cursor reasons about your code. You verify the connection by confirming the tools are available and calling read_creed once.
Step 1: Add a remote MCP server in Cursor
Open Cursor's settings and find the MCP servers section. Add a new remote MCP server and point it at the Creed endpoint:
https://creed.md/mcpCursor supports remote MCP servers, so you are adding Creed as a connector rather than running anything locally. Save the server once the URL is set.
Step 2: Authorize in the browser
Creed MCP uses OAuth, so there is nothing to paste. When Cursor prompts you to authorize, it opens the Creed consent screen in your browser.
- 1.Start the authorization for the Creed server from Cursor's MCP settings.
- 2.A browser opens to the Creed consent screen; sign in to creed.md if needed.
- 3.Click Allow to grant Cursor access to your Creed.
- 4.Return to Cursor; the server should now show as connected.
Step 3: Verify with read_creed
Confirm Cursor can reach your profile before you rely on it.
- Check that the Creed server appears connected and its tools are listed.
- Ask Cursor to call read_creed once and confirm your profile sections return.
Cursor reading your context before it works
The value of connecting Creed shows up before Cursor writes a line. When Cursor reads your profile at the start of substantive work, it plans and edits with your context in hand: your stack from Work, your current direction from Goals, and how you like things done from Preferences. That means fewer wrong assumptions and less time spent correcting output that ignored what you already know.
Because the read happens before planning, the context steers the approach, not just the wording. Cursor is not guessing at conventions you have already settled; it is starting from them.
How updates stay under your control
As Cursor learns something durable about you, it can propose a narrow update to the relevant section rather than rewriting your profile. Updates flow through the creed_* tools and are scoped to specific sections. Whether an edit applies directly or arrives as a proposal depends on your write policy, which an agent can check with get_write_policy before changing anything.
One profile, every tool
The profile Cursor reads is the same one Claude Code, ChatGPT, and your other connected agents read. Your context does not fork per tool. It is plain Markdown you own, with export and delete anytime and no lock-in, and GitHub integration for version control if you want it.
Frequently asked questions
In Cursor's settings, add a remote MCP server pointing to https://creed.md/mcp, then authorize in the browser by clicking Allow on the Creed consent screen while signed in to creed.md.
No. Creed MCP uses OAuth, so authorization happens through a browser consent screen where you click Allow. There is nothing to copy.
Cursor reads your profile before it plans or edits, so your role, stack, goals, and preferences shape its approach from the start rather than after you correct it. You can confirm the read by asking it to call read_creed once.
Only if you allow it. Cursor can check get_write_policy to see whether edits apply directly or as proposals. In proposal mode, updates are suggestions you approve; in direct-edit mode, a trusted agent can apply them through the creed_* tools.