Use your Mind

Portal, MCP and CLI

The Portal is your Mind’s web interface. MCP is the connection that lets a compatible AI tool use the Mind’s advertised tools; a CLI is a command-line program used from a terminal. This page explains which route fits your job and how to connect Claude Code or Codex using your own account, and where to find the other supported tools.

Content updated .

Contents

Use the web portal

Open the Portal for the intended Mind to use chat, browse available information, review requests and manage the settings your account may control. Use chat explains its model connection, sources and saved conversations.

Connecting a personal model for Portal chat is different from connecting an external AI tool to the Mind. A separate model key on the Mind is not required simply to use an external tool you already have.

System mapThree ways to use your Mind

The browser, your agent and the command line all use your Mind and the same access rules.

  1. Choose how to connect

    browser, agent or terminal

  2. Check who is asking

    their account and access

  3. Open allowed areas

    search, read or suggest

  4. Your Mind

    one shared set of files

  5. Activity record

    what happened and when

Get the trusted connection address

In the Portal, open Apps & Tools, then Connect a tool as you. Select your tool and use Copy MCP URL or the setup command shown for that Mind. Wait for the trusted address to load; do not guess it from a hostname.

Each person signs in with their own account. A connection does not inherit the owner’s permissions. An installed program working unattended uses its own app identity, which is a different setup.

Connect Claude Code

Select Claude Code and copy the command supplied by the Portal. Its shape is:

claude mcp add --transport http mind <MCP_URL_FROM_YOUR_PORTAL>

Replace the placeholder with that Mind’s trusted address. Complete browser sign-in as yourself. Follow the Portal’s standing-rule step in the appropriate personal or project instructions so the client knows when to use the Mind. If the account must change, remove this named Mind server and add it again through the supported sign-in flow.

Connect Codex

Select Codex and copy both commands from the Portal. Their shape is:

codex mcp add mind --url <MCP_URL_FROM_YOUR_PORTAL>
codex mcp login mind

Complete browser sign-in with your account. Then add required = true under [mcp_servers.mind] in ~/.codex/config.toml, as the Portal’s current guide instructs. This makes the client wait for the Mind’s tools at startup. If the existing Mind entry has an old static bearer override, remove only that override as the Portal instructs; keep unrelated server settings. Follow its standing-rule step for your personal or project instructions. A token-based identity reported as me is a reason to reconnect, not to mint another personal token.

Connect another tool

The Portal’s connection guide also lists Claude Desktop, ChatGPT, Cursor, Grok and any other remote MCP client that supports OAuth sign-in. Choose yours under Connect a tool as you and follow the steps shown for that Mind.

  • Claude Desktop: add a custom connector under Settings → Connectors, paste the Mind’s URL and complete the browser prompt.
  • ChatGPT: under Plugins, create a custom MCP server with the URL and OAuth authentication. If the form shows STDIO or environment-variable fields, it is the local editor, not this one.
  • Cursor: add an mcpServers.mind entry containing only the URL.
  • Grok: use a custom connector on grok.com. The Portal lists this as a supported alternative; standalone Grok Bot setup is not verified.
  • Other MCP clients: add the URL as a remote server with OAuth. The client must send the resource parameter set to the same URL, or the Mind refuses its token. At sign-in, pick this Mind’s organisation.

Every route ends with the same check: start a fresh conversation, call orient and confirm the person and scopes. Reconnect and sign in gives each tool’s reconnect step.

Verify before using the connection

Start a fresh conversation and ask the client to call orient. Check that the returned Mind, person and scopes are the intended ones. Then read one harmless permitted record. A successful browser sign-in alone does not establish that the client can use the Mind.

For a wrong account, stale browser authorisation or a policy denial, use the named connection’s reconnect flow and have the owner inspect membership or grants where needed. Do not switch to another person’s token to solve a refused read.

Use the supported tools for the job

The actual tools and inputs advertised by the connection are the contract the agent can use. A hosted chat connection may expose fewer management actions than a coding client. If the required action is absent, keep the object reference and hand it to an authorised person or supported client; a similarly named internal endpoint is not a substitute.

Command-line administration is for an authorised operator. Its presence in the software does not give a connected person authority to bypass the Mind’s access or review rules.

Related

Use chat in your Mensara Portal · Ask your first question · How zones and permissions work

Practical guides for people and agents

Start with one useful job

Join the waitlist

Plain text for AI agents. Same content as the human page.

## For agents

- What Mensara is: it gives a team its own Mind, one shared memory for the team and their AI agents, built by Xanadu Labs LLC.
- Status: signup is not open. A person joins the waitlist at https://www.mensara.ai/#waitlist. Agents cannot create an account or sign up.
- How you can help your person: explain what Mensara does, check whether it fits a job their team repeats, and share the waitlist link.
- Agent references: https://www.mensara.ai/llms.txt lists every page as Markdown and the tool input references for connected agents.

---

# Portal, MCP and CLI

Content updated 29 October 2026.

How Mensara works · Use your Mind · [How it works](https://www.mensara.ai/how-it-works)

The Portal is your Mind’s web interface. MCP is the connection that lets a compatible AI tool use the Mind’s advertised tools; a CLI is a command-line program used from a terminal. This page explains which route fits your job and how to connect Claude Code or Codex using your own account, and where to find the other supported tools.

[Page as Markdown](https://www.mensara.ai/how-it-works/portal-mcp-cli.md)

[Operating procedure](https://www.mensara.ai/how-it-works/connect-client-procedure)

## Use the web portal

Open the Portal for the intended Mind to use chat, browse available information, review requests and manage the settings your account may control. [Use chat](https://www.mensara.ai/how-it-works/use-chat) explains its model connection, sources and saved conversations.

Connecting a personal model for Portal chat is different from connecting an external AI tool to the Mind. A separate model key on the Mind is not required simply to use an external tool you already have.

### Three ways to use your Mind

The browser, your agent and the command line all use your Mind and the same access rules.

1. Choose how to connect: browser, agent or terminal
2. Check who is asking: their account and access
3. Open allowed areas: search, read or suggest
4. Your Mind: one shared set of files
5. Activity record: what happened and when

## Get the trusted connection address

In the Portal, open **Apps & Tools**, then **Connect a tool as you**. Select your tool and use **Copy MCP URL** or the setup command shown for that Mind. Wait for the trusted address to load; do not guess it from a hostname.

Each person signs in with their own account. A connection does not inherit the owner’s permissions. An installed program working unattended uses [its own app identity](https://www.mensara.ai/how-it-works/build-an-agent), which is a different setup.

## Connect Claude Code

Select **Claude Code** and copy the command supplied by the Portal. Its shape is:

```
claude mcp add --transport http mind <MCP_URL_FROM_YOUR_PORTAL>
```

Replace the placeholder with that Mind’s trusted address. Complete browser sign-in as yourself. Follow the Portal’s standing-rule step in the appropriate personal or project instructions so the client knows when to use the Mind. If the account must change, remove this named Mind server and add it again through the supported sign-in flow.

## Connect Codex

Select **Codex** and copy both commands from the Portal. Their shape is:

```
codex mcp add mind --url <MCP_URL_FROM_YOUR_PORTAL>
codex mcp login mind
```

Complete browser sign-in with your account. Then add `required = true` under `[mcp_servers.mind]` in `~/.codex/config.toml`, as the Portal’s current guide instructs. This makes the client wait for the Mind’s tools at startup. If the existing Mind entry has an old static bearer override, remove only that override as the Portal instructs; keep unrelated server settings. Follow its standing-rule step for your personal or project instructions. A token-based identity reported as `me` is a reason to reconnect, not to mint another personal token.

## Connect another tool

The Portal’s connection guide also lists Claude Desktop, ChatGPT, Cursor, Grok and any other remote MCP client that supports OAuth sign-in. Choose yours under **Connect a tool as you** and follow the steps shown for that Mind.

- **Claude Desktop:** add a custom connector under Settings → Connectors, paste the Mind’s URL and complete the browser prompt.
- **ChatGPT:** under Plugins, create a custom MCP server with the URL and OAuth authentication. If the form shows STDIO or environment-variable fields, it is the local editor, not this one.
- **Cursor:** add an `mcpServers.mind` entry containing only the URL.
- **Grok:** use a custom connector on grok.com. The Portal lists this as a supported alternative; standalone Grok Bot setup is not verified.
- **Other MCP clients:** add the URL as a remote server with OAuth. The client must send the `resource` parameter set to the same URL, or the Mind refuses its token. At sign-in, pick this Mind’s organisation.

Every route ends with the same check: start a fresh conversation, call `orient` and confirm the person and scopes. [Reconnect and sign in](https://www.mensara.ai/how-it-works/reconnect-and-sign-in#reconnect-a-chat-or-coding-client) gives each tool’s reconnect step.

## Verify before using the connection

Start a fresh conversation and ask the client to call **orient**. Check that the returned Mind, person and scopes are the intended ones. Then read one harmless permitted record. A successful browser sign-in alone does not establish that the client can use the Mind.

For a wrong account, stale browser authorisation or a policy denial, use the named connection’s reconnect flow and have the owner inspect membership or grants where needed. Do not switch to another person’s token to solve a refused read.

## Use the supported tools for the job

The actual tools and inputs advertised by the connection are the contract the agent can use. A hosted chat connection may expose fewer management actions than a coding client. If the required action is absent, keep the object reference and hand it to an authorised person or supported client; a similarly named internal endpoint is not a substitute.

Command-line administration is for an authorised operator. Its presence in the software does not give a connected person authority to bypass the Mind’s access or review rules.

## Related

- [Use chat in your Mensara Portal](https://www.mensara.ai/how-it-works/use-chat)
- [Ask your first question](https://www.mensara.ai/how-it-works/first-useful-question)
- [How zones and permissions work](https://www.mensara.ai/how-it-works/zones-and-permissions)

Previous page: [Team and roles](https://www.mensara.ai/how-it-works/multiplayer)

Next page: [Bring email, calendar events and files into your Mind](https://www.mensara.ai/how-it-works/connections-and-capture)

A connection lets an installed app use an external service you have authorised.

[Join the waitlist](https://www.mensara.ai/?guide_attribution=eyJfcmFpbHMiOnsiZGF0YSI6eyJvcmlnaW5fcGF0aCI6Ii9ob3ctaXQtd29ya3MvcG9ydGFsLW1jcC1jbGkiLCJjdGFfaWQiOiJtYXJrZG93biIsImNvbnRlbnRfcmV2aXNpb24iOiJiOTFmNzRmZTViOGQzMGU0In0sInB1ciI6Imd1aWRlLWNsaWNrIn19--68492f7e018b5d567f1f1f09446764a420915ae3#waitlist)