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

# Set access and safeguards

> Decide who can use a worker, who can configure it, and which supported actions it may perform.

Set boundaries before expanding responsibility
Start here before inviting colleagues, connecting real systems, or preparing a worker for production use.
[Follow the team-access walkthrough →](https://docs.odella.ai/platform/secure-employee-tutorial.md)

## Make three separate decisions

1. **Who can use the worker?** Use teams, roles, and the worker's primary audience. [Access control](https://docs.odella.ai/platform/access-control.md).
2. **Who can change its setup?** Assign a manager and selected editing areas in [Governance](https://docs.odella.ai/workforce/governance.md).
3. **What may it do?** Assign only necessary resources and configure [tool, messaging, and approval policies](https://docs.odella.ai/workforce/settings-security-privacy.md).

Instructions alone do not enforce these restrictions. An approval also does not supply a missing resource permission or provider connection.

## Follow a worked example

> [card] [Restrict a worker to a team](https://docs.odella.ai/platform/secure-employee-tutorial.md)
>
> Create a team and role, assign members, limit editors, and test allowed and
> denied access.

> [card] [Set up a read-only Finance worker](https://docs.odella.ai/platform/finance-read-only-tutorial.md)
>
> Combine provider-side read access, blocked connected writes, and disabled
> outbound messaging.

> [card] [Hand over a worker safely](https://docs.odella.ai/platform/manager-handover-tutorial.md)
>
> Transfer accountability, review unattended work, and remove obsolete access.

> [card] [Review readiness for production](https://docs.odella.ai/platform/security-review.md)
>
> Record your settings, acceptance tests, exceptions, and evidence still
> needed.

## Understand the boundaries before relying on them

| Question | Read |
| - | - |
| How is Odella secured, and what does my team manage? | [Security and data protection](https://docs.odella.ai/platform/security.md) |
| How do capabilities, roles, and resource access fit together? | [Access control](https://docs.odella.ai/platform/access-control.md) and [Roles and permissions](https://docs.odella.ai/platform/roles-and-permissions.md) |
| What do deterministic policies and sandbox isolation cover? | [Tool and browser boundaries](https://docs.odella.ai/platform/tool-and-browser-security.md) |
| Whose signed-in website session is used? | [Browser security](https://docs.odella.ai/platform/browser-security.md) |
| Which user or account connection is selected? | [Credentials and user scope](https://docs.odella.ai/platform/credentials-and-scope.md) |
| Who is accountable for unattended work? | [Manager responsibilities](https://docs.odella.ai/workforce/manager-responsibilities.md) |

Connected-tool approvals are not a universal browser-action policy. Confirm deployment-specific guarantees through Odella's security review rather than assuming every execution path has the same controls.


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