Skip to main content

A clear way to start again

Recover when a coding agent stops

Start a replacement run with a clear choice: begin fresh, or continue with a note about what went wrong.

Make the next run recoverableReview recovery choices

Why this matters

Recovery should create a clear new attempt. It should not quietly rewrite the run that failed.

Before and after

See what changes in the workflow.

Before alterac.ai

  • The task still looks active after the coding-agent process has stopped or disappeared.
  • A replacement agent may repeat work or trust changes that were never submitted.
  • The old agent may submit late and interfere with the new attempt.

With alterac.ai

  • A person chooses Start fresh or Requeue and continue.
  • The replacement gets a new work-run ID and either clean context or a written explanation of what happened.
  • If the old run submits later, alterac.ai rejects its output.

What alterac.ai does

Connect the steps around your coding agents.

alterac.ai creates the replacement run and prevents the old run from taking over the task later. It does not detect or stop the old local process, and it cannot save that process's unsaved edits or terminal state.

  1. 01

    Confirm the run is incomplete

    Use recovery when an Agent-working task no longer has a process you expect to submit.

  2. 02

    Choose how to restart

    Begin fresh from the current task, or explain what happened so the replacement can continue.

  3. 03

    Choose the replacement agent

    Select the coding agent and check its local command again.

  4. 04

    Protect the replacement

    The new run becomes active, and output from the old run can no longer overwrite it.

What stays in your control

Your team keeps the important decisions.

  • Both recovery choices require you to select the coding agent that should run next.
  • An incomplete run kept for context is marked as continued after a problem and left out of work-run metrics.
  • You are still responsible for stopping the old process on its machine or terminal.

A good fit

When to consider this workflow

Choose this workflow when a local agent run becomes incomplete or uncertain and the next attempt needs a safe, visible starting point.

Not a fit

When another approach is better

  • Detecting or stopping remote processes or watching a terminal
  • Automatically saving unsaved local edits
  • Accepting late output from a run that has been replaced
  • General project tracking or work outside software
  • Choosing which coding model, IDE, or coding agent to use
  • Generating code without managing repository work
  • A promise that work will ship faster, code will improve, or conflicts will disappear

Product details and evidence

Learn how it works and see the evidence.

These links explain the product behavior, show the exact limits, and share first-party evidence. Use the Docs guides when you are ready for setup steps and commands.

A clear way to start again

One replacement is queued with the context you chose, and the incomplete run cannot take over the task later.

Create an alterac.ai account.

Make the next run recoverableReview recovery choices