Run a self-hosted agent behind a web proxy

Azure Pipelines | Azure DevOps Server 2019 | TFS 2018 | TFS 2017 | TFS 2015

Note

In Microsoft Team Foundation Server (TFS) 2018 and previous versions, run and release pipelines are called definitions, runs are called builds, service connections are called service endpoints, stages are called environments, and jobs are called phases.

When your self-hosted agent requires a web proxy, you can inform the agent about the proxy during configuration. This allows your agent to connect to Azure Pipelines or TFS through the proxy. This in turn allows the agent to get sources and download artifacts. Finally, it passes the proxy details through to tasks which also need proxy settings in order to reach the web.

Azure Pipelines, TFS 2018 RTM and newer

(Applies to agent version 2.122 and newer.)

To enable the agent to run behind a web proxy, pass --proxyurl, --proxyusername and --proxypassword during agent configuration.

For example:

./config.cmd --proxyurl http://127.0.0.1:8888 --proxyusername "myuser" --proxypassword "mypass"

We store your proxy credential responsibly on each platform to prevent accidental leakage. On Linux, the credential is encrypted with a symmetric key based on the machine ID. On macOS, we use the Keychain. On Windows, we use the Credential Store.

Note

Agent version 122.0, which shipped with TFS 2018 RTM, has a known issue configuring as a service on Windows. Because the Windows Credential Store is per user, you must configure the agent using the same user the service is going to run as. For example, in order to configure the agent service run as mydomain\buildadmin, you must launch config.cmd as mydomain\buildadmin. You can do that by logging into the machine with that user or using Run as a different user in the Windows shell.

How the agent handles the proxy within a build or release job

The agent will talk to Azure DevOps/TFS service through the web proxy specified in the .proxy file.

Since the code for the Get Source task in builds and Download Artifact task in releases are also baked into the agent, those tasks will follow the agent proxy configuration from the .proxy file.

The agent exposes proxy configuration via environment variables for every task execution. Task authors need to use azure-pipelines-task-lib methods to retrieve proxy configuration and handle the proxy within their task.

Note that many tools do not automatically use the agent configured proxy settings. For example, tools such as curl and dotnet may require proxy environment variables such as http_proxy to also be set on the machine.

TFS 2017.2 and older

Important

You also can use this method for Azure Pipelines and newer versions of TFS. We strongly recommend the more modern method, which you can access by switching to the TFS 2018 or Azure Pipelines docs.

In the agent root directory, create a .proxy file with your proxy server url.

echo http://name-of-your-proxy-server:8888 | Out-File .proxy

If your proxy doesn't require authentication, then you're ready to configure and run the agent. See Deploy an agent on Windows.

Note

For backwards compatibility, if the proxy is not specified as described above, the agent also checks for a proxy URL from the VSTS_HTTP_PROXY environment variable.

Proxy authentication

If your proxy requires authentication, the simplest way to handle it is to grant permissions to the user under which the agent runs. Otherwise, you can provide credentials through environment variables. When you provide credentials through environment variables, the agent keeps the credentials secret by masking them in job and diagnostic logs. To grant credentials through environment variables, set the following variables:

$env:VSTS_HTTP_PROXY_USERNAME = "proxyuser"
$env:VSTS_HTTP_PROXY_PASSWORD = "proxypassword"

Note

This procedure enables the agent infrastructure to operate behind a web proxy. Your build pipeline and scripts must still handle proxy configuration for each task and tool you run in your build. For example, if you are using a task that makes a REST API call, you must configure the proxy for that task.

Specify proxy bypass URLs

Create a .proxybypass file in the agent's root directory that specifies regular expressions (in ECMAScript syntax) to match URLs that should bypass the proxy. For example:

github\.com
bitbucket\.com