How Shipfox works

Shipfox automates work for your engineering team with AI agents. You define the work, give agents access to the tools they need, and review the results.

Your team might use Shipfox to build features, review code, investigate problems, or update documentation. Each task can follow its own instructions and checks.

A workflow brings those instructions together. It describes what starts the work, which tasks to perform, and how to check the results.

Choose what starts the workCopy Anchor Link

A workflow can respond to activity in your tools, such as a new ticket or a code update. Shipfox receives that activity as an event. A trigger defines which events start your workflow.

When an event matches a trigger, Shipfox starts a run. A run is one use of your workflow. It includes the status and results of each step, so you can see what finished, what failed, and what the work produced.

Events and triggers explains how you choose what starts the work.

Combine agents, commands, and toolsCopy Anchor Link

A workflow groups tasks into jobs and steps. A job contains steps that Shipfox starts in order. Some steps of a job can run at the same time.

Agent steps use AI to interpret instructions and make decisions. They can read code, make changes, or investigate a question. Command steps perform specific actions, such as running tests or building your project.

You choose which tools each agent can use, such as your issue tracker or code hosting service. Agent access explains how you give agents access to those tools.

Check the resultsCopy Anchor Link

You define checks that decide whether the work can continue. For a code change, these might include tests or formatting checks.

If a check fails, a feedback loop can send the failure details back to the agent. The agent uses those details to try again. The loop stops when the check passes or the agent reaches the attempt limit you set.

You also choose how the workflow shares its results. It might propose code changes for review, add findings to a ticket, or send a message to your team.

Continue the work when new feedback arrivesCopy Anchor Link

Some tasks need more than one round of work. A listening job lets a workflow wait for related events and respond to them.

For example, an agent can respond to a review comment or use new information added to a ticket. You choose which events it responds to and how long it waits.

Follow the workCopy Anchor Link

Each run includes the agent's messages, the tools it used, and the cost of the work. If a step fails, its error details help you understand what went wrong.

Access this information through the dashboard, the API, or MCP tools.

Learn more when you need itCopy Anchor Link

After you complete setup in Shipfox, follow the Quick Start to set up your first workflow with your coding agent. You don't need to read every concept page first.

These pages explain more about how to organize your work:

• Ways to automate engineering work compares Shipfox with local coding agents, agent automations, and CI platforms.
• Workspaces and projects explains how you organize your team and code repositories.
• Runners and execution environments explains where your jobs run.
• Context and templating explains how steps use information from events and earlier results.
• Agents explains when to use an AI agent or a command.
• Agent sessions explains how an agent continues an earlier conversation.
• Workflow concurrency groups explains how you control which runs can work at the same time.