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

# Connect an AI assistant with the Assuric MCP Server

> Connect Claude, ChatGPT or another MCP client to your Assuric workspace to ask questions of your compliance data and, with Editor access, work in your hazard log.

<Note>
  **Quick summary**

  * The Assuric MCP Server, at `https://mcp.assuric.com`, lets an AI assistant answer questions from your Assuric data.
  * You choose *Viewer* (read-only) or *Editor* access when you connect, and the access never exceeds your own Assuric permissions.
  * With *Editor* access, an assistant can work in your hazard log, but asks you to approve each change and never deletes anything.
</Note>

## What MCP is

<Tooltip tip="An open standard for connecting AI assistants to external tools and data.">Model Context Protocol (MCP)</Tooltip> lets an AI assistant talk to a platform directly. Connected to Assuric, it can answer questions from your hazard log, policies and evidence catalogue, without you copying or exporting anything.

<Warning>
  AI can make mistakes. Check answers against their citations, and check every proposed change before approving it.
</Warning>

## Connect your assistant

You may need an administrator of your AI application to add the connector for your organisation.

Each client sends you to the same Assuric sign-in screen. Under **Choose a workspace**, click **Select** beside the workspace to connect. Under **Choose the access to grant**, pick *Viewer* or *Editor* (see [Permissions and access](#permissions-and-access)) and click **Allow**.

<Tabs>
  <Tab title="Claude">
    <Steps>
      <Step title="Open connector settings">
        Go to **Settings → Connectors**, then click **Add custom connector**.
      </Step>

      <Step title="Add the connector">
        Give it a name, for example `Assuric`, and enter the URL:

        ```
        https://mcp.assuric.com
        ```

        Click **Add**, then click **Connect**.
      </Step>

      <Step title="Sign in and choose your access">
        Sign in to Assuric and choose your workspace and access, as above. The tools then appear in your chat.
      </Step>
    </Steps>
  </Tab>

  <Tab title="ChatGPT">
    <Steps>
      <Step title="Turn on developer mode">
        Go to **Settings → Apps → Advanced Settings** and enable *Developer Mode*, then click **Create App**.
      </Step>

      <Step title="Create the app">
        Enter a name, for example `Assuric`, set the connection type to *Server URL*, and enter:

        ```
        https://mcp.assuric.com
        ```

        Leave authentication as *OAuth*, the default, then click **Create**.
      </Step>

      <Step title="Sign in and choose your access">
        Sign in to Assuric when prompted and choose your workspace and access, as above. The tools then appear in your chat.
      </Step>
    </Steps>
  </Tab>

  <Tab title="Other clients">
    Any MCP-compatible client can connect using the same URL and OAuth, including Cursor and others that support remote servers over HTTP.

    **Claude Code:**

    ```bash theme={null}
    claude mcp add --transport http assuric https://mcp.assuric.com
    ```

    Then run `/mcp`, select `assuric`, and authorise in the page that opens.
  </Tab>
</Tabs>

## Permissions and access

| Limit | Set by |
| - | - |
| Your account's existing permissions | Your Assuric role, which the connection can't exceed |
| The access you grant at connect time | You, on the authorisation screen: *Viewer* or *Editor* |

You pick one of 2 access levels:

* *Viewer*, the default: read-only access to your hazards, documents, evidence, organisation context and other records, such as suppliers, risks, incidents, people and products.
* *Editor*: everything *Viewer* can do, plus make changes.

Keep *Viewer* unless you want the assistant to change your hazard log, especially on a shared organisational account.

## Work in your hazard log

With *Editor* access, an assistant can write to your hazard log, for example:

* **Import an existing hazard log:** give it your Excel or CSV file and ask it to create the matching hazards, causes and controls.
* **Draft hazards from a document:** turn a product specification, user story or incident report into proposed hazards and controls.
* **Make bulk updates:** link controls, update statuses or tidy the wording across many hazards at once.

Its limits:

* **You approve first:** it shows each change and writes nothing until you approve. Related changes from one request are approved together, and how approval appears depends on your AI client.
* **No deleting:** it [archives instead](/clinical-safety/managing-hazards#archive-instead-of-deleting), so nothing is lost.
* **Every change is logged:** in **Activity**, against your name, with *MCP* as the *Source* (see [Activity log](/settings/audit-log)).

Then check what it wrote in the hazard log, and score and [approve the log](/clinical-safety/hazard-log-approvals) as usual.

## Working across several workspaces

A connection covers one workspace. For more, add a connector for each and name them to tell them apart, for example `Assuric (My Workspace)`.

<Tip>
  **What we recommend**

  Connect a personal workspace first and try questions you already know the answers to, to calibrate how far to trust it.
</Tip>

## Related articles

<Columns cols={2}>
  <Card title="Add hazards with the Hazard Log Assistant" icon="shield-alert" href="/ask-assuric/hazard-log-assistant">
    Do the same inside Assuric, with Ask Assuric.
  </Card>

  <Card title="Ask Assuric: what it can do" icon="sparkles" href="/ask-assuric/overview">
    Assuric's own AI assistant, working inside the platform.
  </Card>

  <Card title="Documents: Overview" icon="file" href="/documents/key-capabilities">
    What an assistant is reading when it answers.
  </Card>
</Columns>


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