Work with external agents
Use ConsoleX tasks and checks while working in an external AI agent such as Claude Code. Choose a path based on what you already have:
- You have a ConsoleX task: run it through the integration in the external agent, using its saved instructions and supported capabilities.
- You have worked out a method in the external agent: save it as a ConsoleX task, keep the local Skill, and reuse the delivery goal, rules and checks on later runs.
This section covers connecting your computer, running tasks, saving workflows and scheduling. Compare the two paths below, then follow your client's guide. These external applications are separate from the Agents you configure inside ConsoleX.
Two kinds of tasks
Start by checking where the task's method is saved. This determines where it can run and where you change how it works.
| Method in a local Skill | Method in ConsoleX | |
|---|---|---|
| How you create it | Work out a workflow in an external agent, then save it as a task through its ConsoleX integration | Save a ConsoleX Chat as a task, create one from a template, or create one manually in ConsoleX |
| Where the method lives | The main Skill on your computer; ConsoleX keeps a read-only snapshot | The task's instructions and capability selection in ConsoleX |
| Run in ConsoleX | No | Yes |
| Run in an external agent | Yes, with the integration and required main Skill available | Yes, through the integration, using the task's supported capabilities |
| Change the method | Edit the original local Skill | Edit the task in ConsoleX |
Both kinds keep their delivery goal, run rules and checks in ConsoleX. Saving a local workflow as a ConsoleX task does not move its execution to ConsoleX. The task page shows where it runs.
For example, a workflow saved with Claude Code's /consolex:save keeps its method in a local Skill. A task created from a ConsoleX template keeps its method in ConsoleX and can also be started through the external agent's integration.
Which Skills, tools and model does it use?
- Method in a local Skill: the agent uses that Skill and the capabilities its workflow needs. The required Skill and environment must be available wherever the task runs.
- Method in ConsoleX: the integration loads the task's saved instructions, exports supported Skills and provides access to the task's supported, selected ConsoleX tools. Tool calls use your configured ConsoleX connections. Tools or files that need computer permission require Tools and files under Settings → Coding agents.
- Execution model: the external agent uses its own model configuration. The model selected for ConsoleX runs does not configure the external agent.
Not every Skill or tool can be used externally. Availability depends on export support, permissions and connections. The agent may also have its own local capabilities; choosing a ConsoleX task doesn't by itself disable them. See the client's guide for capability limits and isolated execution options.
How a run is checked
- Connect the client to your ConsoleX account on the computer you want to use.
- Select a saved task and start it through the integration. Adding a reusable task Skill is a separate step from running it.
- The external agent does the work and submits the delivery. With automatic checks enabled, ConsoleX checks it against the saved criteria.
- Review the submitted result and report in the task's Run history. By default, failed checks or items needing review are reported for your decision; automatic revision is optional.
Checks use the delivery and evidence the integration can collect. They do not observe everything on the computer. Start with How checks work, then use your client's guide for evidence limits.
Choose where to schedule it
- In ConsoleX: tasks whose method is saved in ConsoleX can use ConsoleX schedules.
- In the external client: use that client's scheduling guide if available. The client starts the work in its own environment; ConsoleX records submitted results and checks. Local runs depend on that computer and application being available.
Connecting an agent doesn't automatically give it a scheduler. Commands, scheduling support and permissions vary by client. Stop an existing schedule before moving the task to another execution environment.
Client guides
Claude Code
| I want to… | Guide |
|---|---|
| Understand the two Claude Code workflows first | Overview — two ways to work |
| Connect this computer to ConsoleX | Install and connect |
| Run an existing task and review its checks | Run and check tasks |
| Reuse a workflow I worked out in Claude Code | Save a session as a task |
| Run a task at a set time on this computer | Schedule tasks in Claude Code |
To create a task in ConsoleX first, see Create a task from a template or Save a chat as a task.