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

# How to connect an AI agent to Jobin with an MCP token

> Connect an AI agent or server-side MCP client to Jobin with a token when it can't complete a browser sign-in, and choose read or write access.

To connect an AI agent to Jobin with an MCP token, create a token in **WorkGroup** > **Integrations** with the provider `jobin-mcp`, then send it as a bearer token to `https://mcp.jobin.cloud/mcp`. Use this for AI agents and server-side MCP clients that cannot complete an interactive browser sign-in.

## What you can do once connected

Your assistant works with your Jobin data in plain language, so you don't switch between tools. For example, you can ask it to:

* **Find and review contacts** by role, location, tag, sequence, or tenure, and read their profiles, notes, and activity.
* **Prepare outreach** by reading a contact's timeline and drafting personalized ideas.
* **Keep records current** by creating contacts, adding tags and notes, and requesting enrichment, with your approval.
* **Work with sequences and pipelines** by finding sequences, adding contacts, moving them between stages, and reviewing performance.
* **Manage follow-ups** by reviewing tasks and creating new ones.

Your connection's permissions decide which actions are available. See the [Jobin MCP overview](/jobin-mcp.md) and the [capabilities reference](/mcp-reference.md).

If your assistant can open a browser sign-in, use its own guide instead:
[ChatGPT](/connect-chatgpt.md), [Claude](/connect-claude.md), or [Codex](/connect-codex.md).

## Before you begin

* An active **Jobin.cloud** account with access to **WorkGroup** > **Integrations**.
* An MCP client that can send a custom `Authorization` header.
* The Jobin MCP connection URL:

```text theme={null}
https://mcp.jobin.cloud/mcp
```

## Create a token

**1. Open Integrations**

In Jobin.cloud, open **WorkGroup** > **Integrations**.

**2. Create a credential**

Create a new integration credential with provider `jobin-mcp`.

**3. Choose scopes**

Choose the MCP scopes to grant. Read access is selected by default; select write access when the
connector should be allowed to make approved changes.

**4. Copy the token**

Copy the returned `mcp_...` token. It is shown once.

**5. Configure your client**

Point your MCP client at `https://mcp.jobin.cloud/mcp` and send the token as a bearer token:

```text theme={null}
Authorization: Bearer mcp_...
```

> [!WARNING]
>
> Treat the token like a password. Store it in your client's secret storage, never in source control or shared chats.

## How tokens are scoped

* A token is tied to the **workgroup where it was created**. It can't be switched to another workgroup;
  create a separate token in each workgroup you need.
* A token only grants the **MCP scopes selected at creation time**.
* Use a dashboard-issued MCP token for the MCP endpoint only.
* Regular API keys (starting with `sk_`) don't work for MCP. Create an MCP token instead.

## Check the connection

Once your client is configured, ask it:

> Check my Jobin.cloud connection.

The response lists the workgroup and permissions the token grants.

## Troubleshooting

> **I lost the token**
>
> The token is shown only once. Create a new integration credential and update your client.

> **Write actions are unavailable**
>
> The token was created with read access only. Create a new token with write access selected and approve actions
> when prompted.

> **The client is rejected**
>
> Confirm the header is exactly `Authorization: Bearer mcp_...` and the URL is `https://mcp.jobin.cloud/mcp`.

> **Results come from the wrong workgroup**
>
> The token stays linked to the workgroup where you created it. Create a token in the workgroup you want to use.

## Next steps

> **[Jobin MCP overview](/jobin-mcp.md)**
>
> What you can do, workgroups, and example prompts.

> **[MCP reference](/mcp-reference.md)**
>
> Every tool, permission, resource, and prompt.


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