Setup MCP

Connect ChatGPT, Claude, Codex, Claude Code, Google Antigravity, or another MCP-compatible client to ClarityLoop.

MCP, or Model Context Protocol, is an open standard introduced by Anthropic in 2024 for connecting AI applications to tools and data. Before MCP, each AI client typically needed a custom integration for each service. MCP gives clients and services a shared way to describe tools, request context, and run actions. The MCP specification defines how clients and servers communicate, including Streamable HTTP and SSE transport.

For ClarityLoop, MCP lets an AI agent discover ClarityLoop tools, understand what each tool can do, and call those tools during a conversation. That is what makes it possible to ask an AI agent to prepare a 1:1, review growth signals, role-play a difficult conversation, explore survey themes, or build a leadership summary using current ClarityLoop context.

Tool calls run as the connected ClarityLoop user, so people, team, and workflow boundaries still apply.

How the Connection Works

You generate a personal MCP API key in ClarityLoop.
Your AI client connects to the ClarityLoop MCP server.
The client scans the tools exposed by ClarityLoop.
During authorization, ClarityLoop asks for your MCP API key.
The client can then call ClarityLoop tools on behalf of that connected user.

For ChatGPT, Claude, Codex, Claude Code, and Google Antigravity, use the OAuth flow when available. You paste the MCP API key into the ClarityLoop authorization screen, not into the chat.

For local or server-side clients that support custom headers, you can send the MCP API key directly in X-ClarityLoop-Api-Key.

Connection Details

Enter these values when your client asks for the ClarityLoop MCP server.

Setting	Value
Server name	`ClarityLoop`
Streamable HTTP endpoint	`https://mcp.clarityloop.com/mcp`
SSE endpoint	`https://mcp.clarityloop.com/sse`
API key header	`X-ClarityLoop-Api-Key: <your_mcp_api_key>`

Start with the Streamable HTTP endpoint unless your client specifically asks for SSE. The /mcp endpoint accepts POST requests.

If you use the API key header, replace <your_mcp_api_key> with your actual key.

Generate an MCP API Key

Each person connects with their own ClarityLoop account. MCP keys are personal, not workspace-wide.

In ClarityLoop, open Workspace > Integrations.
Scroll to Personal Integrations.
Open Personal MCP API Keys.
Generate a new API key.
Copy it immediately and store it safely.

The full key is shown only once. Keys start with clp_mcp_.

Wherever this guide shows <your_mcp_api_key>, replace the whole placeholder with the actual key you generated. Do not paste the literal text <your_mcp_api_key> into your client.

Connect Your Client

Choose the section that matches the AI client your team uses.

OpenAI

ChatGPT

Use this for ChatGPT workspaces that allow custom MCP apps. OpenAI documents the current flow in Developer mode and MCP apps in ChatGPT.

Make sure custom apps or developer mode are enabled for your ChatGPT workspace.
Open Settings > Apps > Create, or, if you are an admin or owner, open Workspace settings > Apps > Create.
Create a new custom app or MCP app.
Enter the server URL:

https://mcp.clarityloop.com/mcp

Choose OAuth authentication if prompted.
Scan tools.
When the ClarityLoop authorization screen opens, paste your MCP API key.
Create the app.
Test it in a chat, then publish or enable it for the intended users.

In ChatGPT Business, Enterprise, and Edu workspaces, an admin or owner may need to enable custom apps or developer mode before users can create and test MCP apps.

Codex

Use this for Codex CLI, the Codex IDE extension, or MCP servers in the ChatGPT desktop app. OpenAI's Codex MCP documentation says these clients share MCP configuration.

In the ChatGPT desktop app, open Settings > MCP servers, add a Streamable HTTP server, and use:

https://mcp.clarityloop.com/mcp

Then authenticate when prompted.

For Codex CLI or project-level configuration, add ClarityLoop to ~/.codex/config.toml or a trusted project .codex/config.toml:

[mcp_servers.clarityloop]
url = "https://mcp.clarityloop.com/mcp"

Then run:

codex mcp login clarityloop
codex mcp list

If you need to use a header instead of OAuth, set the key through an environment variable and reference it from config.toml:

[mcp_servers.clarityloop]
url = "https://mcp.clarityloop.com/mcp"
env_http_headers = { "X-ClarityLoop-Api-Key" = "CLARITYLOOP_MCP_API_KEY" }

Set CLARITYLOOP_MCP_API_KEY to your real MCP API key before starting Codex.

Anthropic

Claude

Use this for Claude chat in the browser, Claude Desktop, Claude mobile, and Claude Cowork. ClarityLoop is a hosted MCP server, so the normal Claude setup is a custom connector.

Anthropic's connector overview lists remote MCP support across Claude.ai, Claude Desktop, Claude Mobile, Claude Code, and Claude Cowork. Their custom connector setup guide says owners add custom connectors for Team and Enterprise workspaces, then each member connects with their own account.

In Claude, open Settings > Connectors.
Choose Add custom connector.
Use the connector name ClarityLoop.
Enter the connector URL:

https://mcp.clarityloop.com/mcp

Add the connector.
Click Connect.
When the ClarityLoop authorization screen opens, paste your MCP API key.
Enable the ClarityLoop tools you want Claude to use.

For Team and Enterprise workspaces, an owner may need to add ClarityLoop under Organization settings > Connectors first. After that, individual users connect it from their own connector settings.

Claude Desktop fallback

Use this only if Claude Desktop cannot connect through Settings > Connectors, or if your organization still manages Claude Desktop MCP servers through local JSON configuration.

This configuration runs mcp-remote on your machine and points it at ClarityLoop's SSE endpoint:

{
  "mcpServers": {
    "ClarityLoop": {
      "command": "npx",
      "args": [
        "-y",
        "mcp-remote",
        "https://mcp.clarityloop.com/sse"
      ]
    }
  }
}

Restart Claude Desktop after changing the configuration. When Claude asks you to authenticate, paste your actual MCP API key into the ClarityLoop authorization screen.

Claude Code

Use this for Claude Code in the terminal. Anthropic documents this in Connect Claude Code to tools via MCP.

If ClarityLoop has already been added in Claude, sign into Claude Code with the same Claude account and run:

/mcp

ClarityLoop should appear in the list of connectors from Claude.

You can also add ClarityLoop directly in Claude Code:

claude mcp add --transport http clarityloop https://mcp.clarityloop.com/mcp

Then run:

/mcp

Follow the authentication flow. When the ClarityLoop authorization screen opens, paste your MCP API key.

If your environment uses direct headers instead of OAuth:

claude mcp add --transport http \
  --header "X-ClarityLoop-Api-Key: clp_mcp_REPLACE_WITH_YOUR_KEY" \
  clarityloop https://mcp.clarityloop.com/mcp

Replace clp_mcp_REPLACE_WITH_YOUR_KEY with your real MCP API key.

Google

Google Antigravity

Use this for new Google agent setups. Google documents MCP setup in the Antigravity MCP guide.

In Antigravity IDE or Antigravity CLI, add a custom MCP server through the MCP server settings, or edit mcp_config.json.

Use serverUrl for Antigravity remote MCP servers.

For OAuth, configure the server URL:

{
  "mcpServers": {
    "clarityloop": {
      "serverUrl": "https://mcp.clarityloop.com/mcp"
    }
  }
}

Then authenticate from Antigravity settings. When the ClarityLoop authorization screen opens, paste your MCP API key.

If your Antigravity setup uses custom headers instead of OAuth, configure the header directly:

{
  "mcpServers": {
    "clarityloop": {
      "serverUrl": "https://mcp.clarityloop.com/mcp",
      "headers": {
        "X-ClarityLoop-Api-Key": "clp_mcp_REPLACE_WITH_YOUR_KEY"
      }
    }
  }
}

Replace clp_mcp_REPLACE_WITH_YOUR_KEY with your real MCP API key.

Gemini CLI

Use this if your team already works with Gemini CLI, especially for existing or enterprise terminal workflows. Google documents MCP server configuration in the Gemini CLI MCP guide.

Add ClarityLoop as an HTTP MCP server:

gemini mcp add --transport http clarityloop https://mcp.clarityloop.com/mcp

Then start Gemini CLI and authenticate:

/mcp auth clarityloop

If your Gemini CLI version or environment expects headers instead of OAuth, configure the server in ~/.gemini/settings.json:

{
  "mcpServers": {
    "clarityloop": {
      "httpUrl": "https://mcp.clarityloop.com/mcp",
      "headers": {
        "X-ClarityLoop-Api-Key": "clp_mcp_REPLACE_WITH_YOUR_KEY"
      }
    }
  }
}

Replace clp_mcp_REPLACE_WITH_YOUR_KEY with your real MCP API key.

Check the connection with:

/mcp list

Debugging Tools

MCP Inspector

MCP Inspector is for testing and debugging. Use it to confirm that the server connects, authorization works, and ClarityLoop tools are visible.

npx @modelcontextprotocol/inspector https://mcp.clarityloop.com/mcp

Complete the authorization flow and paste your MCP API key when prompted.

Other Clients

IDEs and Other Local MCP Clients

If your client supports custom headers, use the Streamable HTTP endpoint with the API key header. This is mainly for local or server-side clients, not browser-based custom apps.

Endpoint:

https://mcp.clarityloop.com/mcp

Header:

X-ClarityLoop-Api-Key: <your_mcp_api_key>

Replace <your_mcp_api_key> with the actual key you generated in ClarityLoop.

Which Connection Method Should I Use?

Use OAuth when the client supports a normal remote MCP authorization flow. This is usually the cleanest path for ChatGPT, Claude, Codex, Claude Code, Antigravity, and MCP Inspector.

Use the API key header when your local or server-side client lets you configure headers directly.

Use /mcp for Streamable HTTP clients. Use /sse for clients or wrappers that still expect SSE.

Advanced Details

Most clients discover OAuth settings automatically from the ClarityLoop MCP server. If you are building or debugging a custom client, the OAuth scope is:

clarityloop:api

After Connecting

Ask the client a simple question first, such as:

What ClarityLoop tools are available?
Show me what context you can access from ClarityLoop.
Help me prepare for my next 1:1 using ClarityLoop context.

Then try a workflow prompt:

Summarize growth signals for my team.
Help me role-play a difficult feedback conversation.
Build a leadership summary from recent feedback, goals, and survey themes.
Suggest dashboard ideas using the ClarityLoop data available to you.

For exposed tools and fields, see the MCP Tools Reference.

For workflow ideas, see MCP Recipes.

Security and key rotation

treat MCP API keys like personal secrets
revoke keys you no longer use
if a key is exposed or lost, generate a new one and update the client

Because access is personal, revoking one key does not affect other people in the workspace.

FAQs

Is an MCP key shared across the workspace?

No. MCP API keys are personal to the account that created them.

What if I lose the key?

Generate a new one and update the client. The original full key cannot be recovered.

Which endpoint should I use?

Use /mcp for direct MCP clients and /sse when the client or wrapper expects an SSE-style remote connection.

Should I edit claude_desktop_config.json for Claude Desktop?

Usually, no. For ClarityLoop's hosted MCP server, try Claude Settings > Connectors first. If that does not work in your Claude Desktop setup, use the Claude Desktop fallback in the Anthropic section.

Which authentication method should I use?

Use OAuth if your client supports it. Use X-ClarityLoop-Api-Key for local or server-side clients that support custom headers but not OAuth.

Can I revoke a key later?

Yes. You can revoke keys from the same Personal MCP API Keys area where you created them.

How the Connection Works​

Connection Details​

Generate an MCP API Key​

Connect Your Client​

OpenAI​

ChatGPT​

Codex​

Anthropic​

Claude​

Claude Code​

Google​

Google Antigravity​

Gemini CLI​

Debugging Tools​

MCP Inspector​

Other Clients​

IDEs and Other Local MCP Clients​

Which Connection Method Should I Use?​

Advanced Details​

After Connecting​

Security and key rotation​

FAQs​

How the Connection Works

Connection Details

Generate an MCP API Key

Connect Your Client

OpenAI

ChatGPT

Codex

Anthropic

Claude

Claude Code

Google

Google Antigravity

Gemini CLI

Debugging Tools

MCP Inspector

Other Clients

IDEs and Other Local MCP Clients

Which Connection Method Should I Use?

Advanced Details

After Connecting

Security and key rotation

FAQs