> ## Documentation Index
> Fetch the complete documentation index at: https://docs.slickerhq.com/llms.txt
> Use this file to discover all available pages before exploring further.

# Connect Slicker to your AI assistant

> Sign in to Slicker from a compatible MCP client, or use an API key

The easiest route is to [copy the setup instruction into your AI assistant](/mcp/overview#get-started-with-your-agent). It asks the assistant to identify your client, guide you through sign-in, and verify the connection. Slicker hosts the server; you do not need to run anything locally.

## Connection details

| Setting | Value |
| - | - |
| Server URL | `https://api.slickerhq.com/mcp` |
| Transport | Remote Streamable HTTP |
| Recommended authentication | Sign in to Slicker with OAuth |
| Alternative authentication | Slicker organization API key, if your client supports a private Bearer token |

Sign in to your Slicker account in the browser. The connection uses the access you authorize during sign-in.

Slicker asks for the `slicker:read` permission to read recovery records, analyze your business, and investigate alerts. Creating or changing alert rules additionally requires `slicker:alerts:write`. Review the permissions shown during sign-in. Your organization's access settings may also restrict who can receive alert-write permission. Even with that permission, review each proposed alert change in your assistant before approving the tool call.

## Codex CLI

```bash theme={null}
codex mcp add slicker --url https://api.slickerhq.com/mcp
codex mcp login slicker --scopes slicker:read
codex mcp list
codex
```

The login command opens your browser for Slicker sign-in. If you also want to manage alerts, reauthorize with both permissions:

```bash theme={null}
codex mcp login slicker --scopes slicker:read,slicker:alerts:write
```

Use `/mcp` inside Codex to check the connection. For an in-depth investigation that exceeds your client's default tool timeout, set `tool_timeout_sec = 900` for this server in your user-level Codex MCP configuration. [Codex's MCP guide](https://learn.chatgpt.com/docs/extend/mcp?surface=cli) covers configuration and OAuth troubleshooting.

## Claude Code

```bash theme={null}
claude mcp add --transport http --scope user slicker https://api.slickerhq.com/mcp
claude mcp login slicker
claude mcp list
claude
```

The login command opens the Slicker sign-in page. You can also sign in through `/mcp` in an interactive Claude Code session. If an existing Slicker entry uses an API-key `Authorization` header, remove that entry and add it again without the header before using OAuth; Claude Code will not fall back to OAuth when an explicit header fails. See [Claude Code's MCP guide](https://code.claude.com/docs/en/mcp) for server and sign-in management.

## Claude on the web or Claude Desktop

Open **Customize → Connectors → Add custom connector**, name it Slicker, and enter the server URL above. Keep the detected OAuth sign-in option rather than choosing **No sign in** or adding a fixed request header. Follow the browser sign-in flow, then enable the connector for your conversation. On a Team or Enterprise plan, an owner may need to add the connector before you can connect your own Slicker account. The [Claude connector guide](https://support.claude.com/en/articles/11175166-get-started-with-custom-connectors-using-remote-mcp) has the current steps and plan requirements.

## ChatGPT

If your plan and workspace allow custom MCP servers, use **ChatGPT on the web → Plugins → Add custom MCP server**. Enter the server URL above, choose **OAuth** authentication, and use **Client ID Metadata Document (CIMD)** when offered. Create and install the connection, then sign in to Slicker when prompted. Do not choose an unauthenticated connection. Availability and controls depend on your ChatGPT plan and workspace; see [OpenAI's custom MCP setup guide](https://developers.openai.com/api/docs/guides/custom-mcp-server).

The ChatGPT desktop app's local MCP configuration is separate from a ChatGPT web plugin. If you use the desktop app's **Settings → MCP servers**, add Slicker as a Streamable HTTP server and select **Authenticate** when prompted. Follow [OpenAI's desktop MCP instructions](https://learn.chatgpt.com/docs/extend/mcp?surface=cli) if your app's labels differ. A web connection does not automatically configure a local Codex session, or vice versa.

## Check the connection

In a new conversation, ask:

> Use only the connected Slicker MCP server. List one recent recovery action as a read-only connection check. If there are no actions, say so. Do not change any alerts.

A successful empty result still confirms that Slicker accepted the connection. Then [explore what Slicker can help with](/mcp/overview#examples-of-what-you-can-do). Seeing a tool listed does not establish that every optional capability is enabled for your organization.

## Organization API key alternative

Slicker also accepts organization API keys for MCP clients that can privately supply an `Authorization: Bearer` header. Create or retrieve a key at [Slicker API keys](https://auth.slickerhq.com/org/api_keys). **An organization API key is a full administrator credential**, unlike a scoped OAuth connection. Do not paste it into an AI chat, a command argument, source code, or a shared connector.

For Codex CLI, put the key in a local environment variable before starting the client, then configure only the variable's **name**:

```bash theme={null}
printf 'Slicker API key: '
read -r -s SLICKER_API_KEY
printf '\n'
export SLICKER_API_KEY
codex mcp add slicker-key --url https://api.slickerhq.com/mcp --bearer-token-env-var SLICKER_API_KEY
codex
```

Supply the variable again in each new terminal session or load it from your secret manager. Do not use the same server name for both the OAuth and API-key connections. Other clients need their own private Bearer-token setting; ChatGPT web's custom MCP plugin flow does not accept an organization API key.

## Troubleshooting

| Symptom | What to check |
| - | - |
| Sign-in succeeds, but the client says `authentication required` | Refresh or restart the client's MCP connection and verify you signed in to the server URL above. If the connection still fails, contact Slicker support. Do not paste an access token into chat. |
| ChatGPT reports an OpenID Connect conflict during setup | Turn off the optional OpenID Connect setting for this MCP connection and use OAuth sign-in. |
| Recovery records work, but alert changes are denied | The OAuth connection may lack `slicker:alerts:write`, or your Slicker role may not be allowed to grant it. Reauthorize with the write permission if appropriate. |
| Recovery actions work, but business questions do not | Your organization may not have AI analytics data access enabled. Contact your Slicker administrator or Slicker support. |
| Alert tools are unavailable | Proactive Alerts may not be enabled for your organization. |
| A preset or requested custom signal is unavailable | Your connected data may not support that metric or segment. Ask the assistant to explain the limitation and suggest an available alternative. |
| An analysis has little or no data | Check the period and connected data. Ask the assistant to state the exact dates and any coverage gaps. |

For help, [contact Slicker support](mailto:support@slickerhq.com). See also [Slicker API authentication](/api-reference/overview#authentication).


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.