Add a custom connector to Claude (free plan included)

Updated 2 October 2026. Steps checked on claude.ai on 1 October 2026. Claude's menus change: if a label has moved, the principle stays the same.

To add a custom connector to Claude, open claude.ai (web or desktop app), go to Settings → Connectors, click “+” then “Add custom connector”, enter a name and the MCP server URL, click Add, then Connect to authorize access with OAuth. It works on every plan: the free plan allows one custom connector; Pro, Max, Team and Enterprise allow several.

This guide uses Pryzm Memory as the example: a persistent memory shared across several AIs over MCP. The steps are the same for any remote MCP server. A section at the end covers Claude Code, Anthropic's command-line tool.

Custom connector: what is it?

A connector lets Claude call tools hosted elsewhere: search a service, read from it or save data to it. Anthropic offers ready-made connectors; a custom connector is one you add yourself, from the URL of an MCP server (Model Context Protocol, an open protocol).

Nothing to install: the server runs at its publisher, and Claude connects to it over the internet. You stay in control of access, since the connection goes through OAuth and can be removed at any time.

What you need

• A Claude account, on any plan: Free (one custom connector), Pro, Max, Team or Enterprise.
• claude.ai in a browser, or the Claude desktop app. Connectors are added from either of these.
• The MCP server URL. For Pryzm: https://pryzm-memory.com/mcp.
• An account with the service you connect, for the authorization step. Pryzm is free up to 100 memories.

The steps on claude.ai

1. Open settings. On claude.ai or in the desktop app, open your profile menu, then Settings.
2. Go to Connectors. That's where your existing connectors are listed.
3. Add the connector. Click “+”, then “Add custom connector”.
4. Fill it in. Give it a name (for example “Pryzm”), paste the MCP server URL, then click Add.
5. Sign in. Click Connect: the service's sign-in page opens. Sign in to your account to authorize Claude. It receives its own access token, which you can revoke at any time.
6. Check. The connector shows as connected. In a chat, its tools are available from the tools menu.

To test it, ask a question that forces Claude to use the connector, for example: “Search my memory for what you know about my current projects.” Claude shows the tool call before answering, and may ask for your permission the first time.

Free plan: one connector, chosen well

On the Free plan you get one custom connector. If you want a second one, you need to remove the first or move to a paid plan. For a memory like Pryzm, one is enough: the same connector searches, saves and gives project overviews.

On Team and Enterprise, connector management may depend on your organisation: if the option is missing, an owner or admin may need to allow it first. Each member then signs in with their own account.

Get Claude to check the memory on its own

A connector only helps if Claude thinks of using it. To make it check the memory at the start of every conversation and save what matters without being asked, add an instruction in Settings → General, in the field “What personal preferences should Claude consider in responses?”.

For Pryzm, the ready-to-paste text is on the connection page, in the “main memory” section.

Claude Code: a single command

In Claude Code, Anthropic's command-line tool, a remote MCP server is added from the terminal:

claude mcp add --transport http pryzm https://pryzm-memory.com/mcp

• Authorize. Start Claude Code, type /mcp, pick the “pryzm” server and follow the OAuth sign-in that opens in your browser.
• Check. The command claude mcp list should show “pryzm” among the servers.
• Add the instruction. Paste the “main memory” instructions into ~/.claude/CLAUDE.md so they apply to all your projects.

Claude Code comes with the plans that include it; the command above is the same whatever the plan.

Limits worth knowing

• One custom connector on the free plan. No more.
• Menus change. Anthropic updates the interface regularly; exact button labels may differ from this guide.
• You are responsible for the server you plug in. A custom connector is not reviewed by Anthropic the way a connector from its directory is: only connect servers whose publisher you know.
• The server sees what Claude sends it. At Pryzm, each account is encrypted with its own key (AES-256-GCM), with servers in Germany; but during a session, the server decrypts your memories in memory so it can search them. It is not end-to-end encryption.

If something doesn't work

• The add button is greyed out on the free plan: you probably already have a custom connector. Remove it or use it.
• Error after Add: check the URL, with no extra space or trailing slash.
• OAuth sign-in fails: allow pop-ups for claude.ai and check you are signed in to the right account on the service.
• The connector is connected but Claude doesn't use it: check it is enabled in the chat's tools menu, then add the instruction to your preferences.

When it isn't useful

If you only use Claude, its built-in memory and projects are probably enough. An external memory like Pryzm helps when you also work with ChatGPT, Perplexity, Cursor or Gemini CLI, and want them to share the same context.

For ChatGPT, see how to connect ChatGPT to an external memory. Guides for every AI are on the connection page.