Give your agent real workflows to work on.

Connect your coding agent to Tilebox. Give it the tools to find datasets, write workflow code, run test jobs, and inspect what happened.

Your instruction

Find the available scenes. Build a processing workflow. Show me the changes before deploying.

I’ll inspect the datasets and workflow tools first. Then I’ll prepare the code for review.

Ready for your reviewWorkflow changes
  • tasks.pyScene-processing task
  • runner.pyRegistered task
  • tilebox.workflow.tomlDevelopment target
Next: review, publish, and test on a dev cluster.
Illustrative agent conversation and code review.

From a question to a workflow you can inspect.

Take an idea beyond generated code. Your agent can work with real datasets and executions, then use the results to make the next change.

  1. Discover the starting point

    The agent can inspect available commands, load Tilebox skills, and query datasets. Start with a read-only request so you can check its context before it changes resources. Explore the data layer.

  2. Turn the idea into code

    Ask the agent to write a workflow around your analysis. Review the changes, then have it publish and deploy a release to your development environment.

  3. Inspect, fix, and run again

    Have the agent inspect a test job’s logs and traces. It can use that evidence to propose a fix, test the change, and help you check the output.

See the workflow iteration guide

Tools in the terminal. Context for the task.

The CLI is the default for coding agents. It works where the agent already edits files and runs tests, with structured output, command discovery, and file-based inputs.

Tilebox skills teach the agent how to combine those tools for dataset management, workflow authoring, and job monitoring.

Using a chat tool without terminal access? The MCP server is the alternative interface.

Your coding agentWorks in your repository
Tilebox skillsTask-level instructions
Tilebox CLIDiscover commands · run actions · read output
DatasetsReleasesJobs & logs
Skills guide the work. Authenticated tools act on Tilebox resources.

Give it access.
Keep your judgment.

Ask the agent to inspect first. Review its code and intended resource changes before authorizing them.

Test changes in a development environment before using them in production. Check the analysis output as well as whether the job completed.

You decide where runners execute and when work reaches production. Set your review and approval rules in the agent tools your team uses.

See how workflows run or choose the execution environment.