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

# Integrations and credentials

> Let your Agent call external services without ever seeing their secrets, control which credentials it may use, and attach other Tela resources as tools

An Agent often needs more than the files it receives: a search in a case-law database, a lookup in your CRM, a Canvas you already trust. **Integrations** give the Agent external services it can call, with the credential kept out of its reach. **Agent Credentials** control which of your Workspace credentials it may use. The **Capabilities** tab attaches published Canvases, Workflows and Agents as tools.

For the contract assistant of the [running example](https://docs.tela.com/en/agent/overview.md#the-running-example), an integration named `caselaw` points to the legal database the team subscribes to, so the Agent can check how a clause has been interpreted.

## Integrations

The **Integrations** section of the sidebar lists the external services this Agent can call, with the **Service** name, the **Header** used and the **Credential** that authorizes it. The credential stays in the Workspace and is applied by Tela at request time; the Agent never receives the value.

Only Workspace administrators can add or edit integrations. Other members see the list in read-only mode, including a **Credential not found** badge when the credential an integration relies on no longer exists.

![Integrations section listing the caselaw integration with its header and credential](https://mintcdn.com/meistrari/IBJhAp-onVS37l80/images/agent/integrations.png?fit=max\&auto=format\&n=IBJhAp-onVS37l80\&q=85\&s=74f3259c802cc3e3151a24ae3548b618)

### Adding an integration

**1. Open the dialog**

Click **New integration**. The **Integrations** dialog opens with one empty integration; click **Add integration** at the bottom for more.

**2. Name the service**

Under **Name**, type how the Agent will refer to the service: lowercase letters, numbers and hyphens, for example `caselaw`. The names `vault`, `tela`, `doc` and `otel` are reserved, and names must be unique in the Agent.

**3. Set the Base URL**

Under **Base URL**, type the root address of the service, for example `https://api.example.com`. It must use `https`.

**4. Choose how the credential is sent**

Under **Authentication**, choose **API key header** or **Bearer token**. For an API key, you can set the **Header name (optional)**, for example `x-api-key`.

**5. Pick or create the credential**

Under **Credential**, select a Workspace credential (listed as `KEY · workspace`) or one that belongs to this Agent (`KEY · this agent`). Or choose **+ Create a new credential** and paste the value: it is stored encrypted, scoped to this Agent, under a key Tela derives from the name, such as `CASELAW_API_KEY`.

**6. Save**

Click **Save**. The integration appears in the table.

The dialog states the rule: *Credentials are stored encrypted and applied by Tela's proxy at runtime; the raw value never reaches the agent's sandbox.* Two details follow from it:

* If you change the **Base URL** after choosing a credential, you must select (or re-enter) the credential again to authorize it for the new address.
* Credentials the Agent already uses as plain secrets (see below) are not listed: an integration can't share a key with them.

### Telling the Agent to use it

In the Purpose, name the service and say when to use it, in plain language: *"When a clause mentions a statute, look it up in the caselaw integration before flagging it."* The Agent calls the service by name; Tela adds the credential on the way out.

> [!NOTE]
>
> Which paths the Agent calls under the Base URL depends on its instructions and on the service's API, so describe the endpoints it should use in the Purpose or in a skill. Test the integration in the **Test your agent** panel and read the **Debug** view to see the calls it made.

## Agent Credentials

Integrations cover services with an API key or a token. For everything else the Agent may need as a secret, Tela uses the same [Credentials](https://docs.tela.com/en/credentials.md) as Workflows: the Agent refers to a credential by name and never sees its value.

Administrators open them from the `…` menu next to the Agent title, under **Agent Credentials**. The dialog has two tabs:

* **Agent Credentials**: credentials specific to this Agent. They override Workspace credentials with the same key, which is useful when this Agent must use a different account than the rest of the Workspace.
* **Usage Permissions**: the **Credential Access Control**. Toggle which credentials this Agent is allowed to use; disabled credentials are not available during execution. With nothing toggled, all credentials are enabled.

> [!WARNING]
>
> When you restrict the list, include every credential the Agent needs. A credential that is not allowed behaves as if it did not exist and the call that needs it fails.

To tell the Agent which credential to use, name it in the Purpose, for example *"Use the SLACK\_BOT\_TOKEN credential to post the summary in the #legal channel"*. The naming rules and the rotation flow are the same as in [Credentials](https://docs.tela.com/en/credentials.md#managing-credentials).

## Other Tela resources as tools

A published Canvas, Workflow or Agent of your Workspace can become a tool your Agent calls. This is how you reuse a Canvas that already extracts contract parties reliably, or a Workflow that files a document in your systems, instead of teaching the Agent to do it from scratch.

Open the **Skills** dialog (click **Add skill** in the **Skills** section) and switch to the **Capabilities** tab. Pick a Project in the dropdown, search if needed, and hover a card to click **Add**. Each card shows the resource title and its kind: **Canvas**, **Workflow** or **Agent**. Only published resources are listed, and the current Agent is not offered to itself.

Once added, describe in the Purpose when to call it, the same way you would for a skill. The Agent fills the resource's inputs from what it knows and receives the published result.

> [!NOTE]
>
> A Lightweight Agent has no **Skills** dialog. There, click **Add tool** in the **Tools** group instead, pick a published Workflow or Canvas and, optionally, write a description. The Agent uses the inputs of the published resource to decide what to send. See [The Lightweight Agent page](https://docs.tela.com/en/agent/purpose-and-model.md#the-lightweight-agent-page).

![Capabilities tab of the Skills dialog listing published Canvases, Workflows and Agents of a Project](https://mintcdn.com/meistrari/IBJhAp-onVS37l80/images/agent/capabilities.png?fit=max\&auto=format\&n=IBJhAp-onVS37l80\&q=85\&s=3cb97592e66a14e3a391e47962b45217)


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