Marcora

official

Your team's brand context in any AI tool. Marcora connects AI assistants to your Brand Foundation and Reference Library.

View on marcora.ai

What can you do with Marcora MCP?

  • Create blueprints — Ask your AI assistant to create blueprints inside Marcora without leaving the chat.
  • Generate content — Have the assistant generate content using your Marcora tools and context.
  • Manage context — Ask the assistant to manage your Marcora context items during a conversation.
  • Connect via URL — Point any MCP client at https://mcp.marcora.ai using OAuth or an API key as the bearer.
  • Use the workflows skill — Install the Marcora AI Workflows skill so your client follows Marcora's object model and generation patterns.

Hosted MCP Server

npx add-mcp 'https://mcp.marcora.ai'

Installs into Claude Code, Codex, Cursor and more

Documentation

Marcora MCP Server

The Marcora MCP Server exposes all of your Marcora tools directly inside AI assistants like Claude and ChatGPT, letting you create blueprints, generate content, manage context, and more — without leaving your AI chat.

Connection URL

https://mcp.marcora.ai

This is the single URL for connecting to the Marcora MCP Server. It works both with OAuth — ideal for Claude and ChatGPT — and with a direct API key or token, supplied as the bearer. Transport is negotiated automatically (Streamable HTTP, with SSE fallback), so the same URL works across MCP clients.

You can create and manage your MCP API keys in Integrations Settings — then use your API key with the URL above.

Connecting in Claude

Marcora is listed in Claude's connector directory, so there's no URL and no token to paste. That's fewer steps than adding it by hand — not zero: you still approve access, and Claude asks permission the first time it uses each tool.

  1. Open the Marcora listing in Claude's connector directory.
  2. Click Connect. Marcora's authorization page opens — click Allow. (If you already have MCP tokens, you'll be asked which one to use first.) Claude returns you to a new chat.
  3. Just ask. Marcora is already switched on in every new conversation — there is nothing to enable in the composer. Try: "What's my brand foundation?"
  4. Approve the tool the first time Claude reaches for it. Claude pauses and asks — "Claude wants to use Get Brand Foundation from Marcora: Brand Context & Writing" — with Always allow and Deny. Click Always allow.

⚠️ Step 4 is the one people miss. Until you click Always allow, nothing reaches Marcora: Claude will have found the tools but never actually called one. The prompt is per tool, not per connector, so expect it again the first time each new Marcora tool comes up. Approvals then persist across all your conversations. (Allow once, under the small chevron, approves a single call and asks again next time.)

Claude labels Marcora a Community connector, meaning it passed Anthropic's automated checks rather than their in-depth review. That describes how the listing was assessed — it doesn't limit what Marcora can do once you connect it.

Managing the connection later: go to Customize → Connectors (claude.ai/customize/connectors), where Marcora appears as Connected. Note that it only shows up in that list after you connect — searching for it there beforehand returns nothing, which is why step 1 starts at the directory.

Adding Marcora by URL instead

The directory is a catalogue, not a different kind of connector — adding Marcora by URL gives you exactly the same tools. Use this route if you're adding Marcora for a whole Team or Enterprise workspace, or if you'd simply rather paste the URL.

  1. In Claude, go to Customize → Connectors, click Add, then choose Add custom connector.
  2. Give the connector a name and paste your Marcora MCP Server URL: https://mcp.marcora.ai
  3. Click Add, then click Connect next to "Marcora" and approve access to finish the OAuth flow.

Then carry on from step 3 above — the per-tool approval prompt works the same way.

On the Free Claude plan you can have only one custom connector. Connecting from the directory doesn't involve adding a custom connector at all, so it leaves that slot free.

Connecting in ChatGPT

Marcora has a listing in ChatGPT's plugin directory, so there's no URL and no token to paste. That's fewer steps than adding it by hand, not zero: you still sign in to Marcora and approve access.

  1. Open the Marcora listing in ChatGPT's plugin directory.
  2. Click Install plugin. Sign in to ChatGPT first if it asks.
  3. Marcora's sign-in page opens. Log in with your Marcora account and approve access. ChatGPT returns you to your chat.
  4. Just ask. Start a new chat and try: "What's my brand foundation?" ChatGPT uses installed plugins automatically when they're relevant. To call Marcora directly, type @Marcora, select + and enter its name, or choose Plugins in the composer. ChatGPT may ask you to confirm an action before it runs.

Who can install it: the plugin directory is available across ChatGPT plans, but whether you can install a given plugin depends on your plan, workspace, role, and region. In a managed workspace an admin may need to allow Marcora. Business workspaces have apps on by default, while on Enterprise and Edu an admin may need to enable it first. If Marcora shows Disabled by admin, ask your workspace admin to allow it. See OpenAI's guide: Plugins in ChatGPT.

Managing the connection later: open Plugins in ChatGPT and select Installed, where Marcora appears.

Adding Marcora as a custom MCP server instead

Adding Marcora by URL gives you exactly the same tools. Use this route if you're adding Marcora for a whole workspace, or you'd simply rather paste the URL. Full custom MCP server support is available on ChatGPT Business and Enterprise/Edu, on the web, and on Business only workspace admins and owners can create them. Pro can connect custom MCP servers with read and fetch permissions only.

  1. In ChatGPT, go to Plugins, select the plus button, then choose Add custom MCP server. You don't need to turn on Developer mode first.
  2. Give it a name. Under Connection, enter your Marcora MCP Server URL: https://mcp.marcora.ai. Set authentication to OAuth.
  3. Review the risk warning and select I understand and want to continue, then select Create as a plugin and sign in to Marcora to finish the OAuth flow.

Then carry on from step 4 above. Older workspace setups use Settings > Apps > Create instead, and OpenAI is still changing this area, so check Connecting custom MCP servers in ChatGPT if your screen looks different.

The Marcora AI Workflows Skill (recommended)

The Marcora AI Workflows Anthropic Skill ships alongside this server. It teaches your AI client Marcora's mental model and the right patterns for every task — the object model, content generation, blueprints, projects, briefs, context items, the content plans board and playbooks, and building reusable multi-step workflows — so the agent stops misfiring on tasks that look ambiguous from tool descriptions alone.

  • Marcora AI Workflows (marcora-mcp, v0.7.18) — object model (content, blueprints, projects, context items), content-generation patterns, the 4-layer context model, project-brief mechanics, the content plans board and playbooks, and building reusable multi-step workflows. Your AI client picks it up on demand based on the task.

Download

Get the .skill file from the latest release:

  • marcora-mcp.skill — the Marcora skill (a zip archive containing SKILL.md and its bundled reference files).

Install

Claude Code:

mkdir -p ~/.claude/skills
unzip ~/Downloads/marcora-mcp.skill -d ~/.claude/skills/

Restart Claude Code. The skill loads automatically on any Marcora-related task.

Claude Desktop:

Import the .skill file via Claude Desktop's Skills folder. See Anthropic's Skills documentation for the current path on your platform.

ChatGPT, Cursor, and other non-Anthropic clients:

These don't directly support Anthropic Skills (yet). Workaround: unzip the .skill file and paste the contents of SKILL.md into your client's custom instructions or system-prompt slot. Bundled reference files can be pasted in too if your client supports longer context.

Power-user install (clone and symlink)

If you'd rather track main directly:

git clone https://github.com/ccromp/marcora-mcp.git
ln -s "$(pwd)/marcora-mcp/skill/marcora-mcp" ~/.claude/skills/marcora-mcp

Releases

Releases are tagged marcora-vX.Y.Z — see Releases. Each release attaches the bundled skill as marcora-mcp.skill. The skill's own version is the metadata.version field in its SKILL.md, and it moves independently of the release tag.

On this page