Skip to main content
Version: 2.0.x

Local backend


The local backend will execute the pipelines on the local system without any isolation of any kind.


This backend is still pretty new and cannot be treated as stable. Its implementation and configuration can change at any time.

Since the code runs directly in the same context as the agent (same user, same filesystem), a malicious pipeline could be used to access the agent configuration especially the WOODPECKER_AGENT_SECRET variable.

It is recommended to use this backend only for private setup where the code and pipeline can be trusted. You shouldn't use it for a public facing CI where anyone can submit code or add new repositories. You shouldn't execute the agent as a privileged user (root).

The local backend will use a random directory in $TMPDIR to store the cloned code and execute commands.

In order to use this backend, you need to download (or build) the binary of the agent, configure it and run it on the host machine.



Enable connection to the server from the outside of the docker environment by exposing the port 9000:

# docker-compose.yml for the server
version: '3'

- 9000:9000
- [...]


You can use the .env file to store environmental variables for configuration. At the minimum you need the following information:

# .env for the agent

Running the agentโ€‹

Start the agent from the directory with the .env file:



When using the local backend, the plugin-git binary must be in your $PATH for the default clone step to work. If not, you can still write a manual clone step.

Further configurationโ€‹

Specify the shell to be used for a pipeline stepโ€‹

The image entry is used to specify the shell, such as Bash or Fish, that is used to run the commands.

# .woodpecker.yml

image: bash
commands: [...]

Plugins as Executable Binariesโ€‹

image: /usr/bin/tree

If no commands are provided, we treat them as plugins in the usual manner. In the context of the local backend, plugins are simply executable binaries, which can be located using their name if they are listed in $PATH, or through an absolute path.

Using labels to filter tasksโ€‹

You can use the agent configuration options and the pipeline syntax to only run certain pipelines on certain agents. Example:

Define a label type with value exec for a particular agent:

# .env for the agent


Then, use this label type with value exec in the pipeline definition, to only run on this agent:

# .woodpecker.yml

type: exec

steps: [...]

Change temp directoryโ€‹

We use the default temp directory to create folders for workflows. This directory can be changed by: