Connect Reeve to Claude

Reeve exposes a remote MCP server at:

https://api.meetreeve.com/mcp

In Claude Desktop and claude.ai, add this URL as a custom MCP connector. In Claude Code, install the Reeve plugin instead — it connects Reeve for you and adds automatic memory. Authentication uses OAuth everywhere: you sign in to your Reeve account once, and Claude handles the token from there. No manual token pasting required.

Open Claude Desktop settings

Go to Settings → Developer (or Integrations, depending on your version) and find the MCP / custom connectors section.

Add a custom connector

Click Add custom connector (or Edit config to add to claude_desktop_config.json directly). Add the Reeve server URL:

{
  "mcpServers": {
    "reeve": {
      "url": "https://api.meetreeve.com/mcp"
    }
  }
}

Save the config and restart Claude Desktop.

Authorize Reeve

When you first use a Reeve tool, Claude Desktop will prompt you to sign in to Reeve. Complete the OAuth flow in your browser — you'll be redirected back once connected.

For the current Claude Desktop connector UI, see Anthropic's MCP documentation.

Open Integrations

In claude.ai, go to Settings → Integrations (or Connectors).

Add a custom connector

Click Add custom connector and paste the Reeve MCP server URL:

https://api.meetreeve.com/mcp

Click Connect (or Authorize). You'll be taken to Reeve's OAuth sign-in. Log in with your Reeve account and approve the connection. You're redirected back to claude.ai once authorized.

The exact connector UI may vary as Anthropic updates claude.ai. See Anthropic's connector documentation for the current steps.

In Claude Code, install the reeve plugin. It connects Claude Code to Reeve's full tool surface and turns on automatic memory — no API key or local setup needed.

Add the Reeve marketplace

In Claude Code, run:

/plugin marketplace add MindFortressInc/reeve-claude-plugin

Install the plugin

/plugin install reeve@reeve

Run /mcp, select plugin:reeve:reeve, and complete the browser sign-in with your Reeve account. You only do this once.

Until you sign in, the memory hooks can't reach Reeve: Claude Code may show a hook error notice, and your session continues normally.

What automatic memory does

Recall. Before each prompt, the plugin asks Reeve for memory relevant to what you just typed and adds it to that turn's context. If nothing is relevant, nothing is added.
Capture. Each turn — the prompt you submitted and Claude's final reply, plus the session id and working directory — is saved to your Reeve memory. Tool calls, tool output, file contents, and Claude's intermediate messages are not captured.
Privacy. Captured turns go to your personal Reeve memory only, under the account you signed in with — never to an organization or team scope. The full text of your prompts leaves your machine, so don't paste secrets into prompts while the plugin is enabled.
Opt out. Turn off automatic memory writes in your Reeve memory settings (auto_write_enabled), or disable the plugin from /plugin to stop the hooks entirely.
Uninstall. Run /plugin uninstall reeve@reeve.

See the plugin README for exactly what is sent and when. Your claude.ai Reeve connector can stay on alongside the plugin.

Added Reeve with claude mcp add before? Remove that entry (claude mcp remove reeve) and use the plugin. Keeping both gives Claude Code two copies of Reeve's tools, and because manually added servers take precedence over plugin servers, a manual server at the plugin's URL would hide the plugin's server and leave the memory hooks with nothing to call.

Alternative: tools without automatic memory

If you only want Reeve's tools, with no recall or capture, register the server directly instead of installing the plugin:

claude mcp add --transport http reeve https://api.meetreeve.com/mcp

On first use, you'll be prompted to complete the OAuth sign-in in your browser. Use this or the plugin, not both — the plugin already connects Reeve, so don't add this server alongside it.

See Claude Code MCP documentation for more on managing MCP servers in Claude Code.

What happens after you connect

Once connected and authorized, Reeve appears as an available integration in Claude. You can start a conversation and ask Claude to work with your Reeve workspace — no extra setup needed. Claude discovers what's available by calling get_context, then browses and calls tools as your request needs them (see The tools for how that discovery works).

To verify the connection is working, ask Claude: "What brands do I have in Reeve?" — Claude will call get_context and return your workspace details.

Memory. In Claude Code, the plugin recalls and captures memory automatically (see the Claude Code tab above). Claude Desktop and claude.ai have no hooks, so nothing is captured automatically there. Instead, Reeve's server instructions tell Claude to save durable facts — preferences, corrections, decisions — with memory_write as it goes, and Claude can look them up with memory_search. That's a weaker guarantee than the plugin's per-turn capture, so if something matters, ask Claude to remember it.

Advanced: programmatic access

For programmatic or substrate consumers (e.g., automated workflows not tied to a user account), the Reeve MCP also supports service-level authentication via an X-Reeve-Host-Key header on /mcp. A host-key connection sees a flat enumeration of every tool its capability grants allow, rather than the progressive get_context → list_tools → describe_tools → use flow OAuth connections use. This is an internal/advanced path — contact support if you need it.

Connect Reeve to Claude

On this page