Skip to main content
Project autopilot creates a single dedicated automation for an entire project. When it runs, Neo reads all previous task results and project settings, then determines what to do next. Each run creates a new task in the project and is recorded in the Execution History panel on the Autopilot tab. Autopilot is different from task automations. A task automation runs a specific prompt you write, on a schedule you set, for a repeatable job. Autopilot uses a fixed, system-generated prompt and draws on everything the project already knows to decide what to work on.

How it works

When autopilot runs, Neo pulls all task context from the project, including completed tasks, working memory (facts, insights, todos, and tracked files), and the project settings. It then creates a new task using the accumulated project context. Every task completed inside the project, including those created by previous autopilot runs, becomes part of the context for the next run. The more work that has been done in the project, the more informed each subsequent run becomes. Autopilot adds one new task per run. It does not automate, re-run, or take over the tasks already in the project, and enabling it does not change how the tasks you start yourself behave. The autopilot prompt is fixed and cannot be customized. Every run starts from the same standing instruction, which is to continue making progress on the project from its latest context. Nothing new is written or generated for each run, and there is no prompt field to fill in or edit. What changes between runs is the context wrapped around that instruction: the project settings as they read at run time, and everything the project has accumulated since the last run. That is why two runs of the same autopilot behave differently while the instruction stays identical. If you need to control exactly what runs on a schedule, use a task automation instead.

Steering autopilot

Because there is no prompt to write, project Settings is where you steer autopilot. The scope, objectives, focus areas, known findings, and constraints you record there are read before every run, so they do the job the prompt would otherwise do. Autopilot is most useful on a project that already has history behind it. Run a few tasks yourself, fill in Settings, then enable it, so the first automated run has real context to build on instead of starting from an empty project.

Enabling autopilot

Open the Autopilot tab inside a project and configure the schedule.
1

Set the frequency

Choose Daily, Weekly, Monthly, or Custom. Custom lets you select the specific days of the week autopilot runs, which is how you keep runs on working days only.
2

Set the run time

Pick the hour and minute the run starts. Run times are set in UTC, in fifteen minute increments.
3

Choose a reporting mode

Changes only reports on a run when it finds something new compared with the previous run. Full report returns a complete report of every run regardless of what changed.
Click Enable Autopilot to activate the schedule. Neo runs at the next scheduled time using the full accumulated project context. Autopilot does not expose model, agent, secret, or file selection. Runs use your workspace defaults and Neo selects agents automatically. If you need any of those pinned to a specific value, use a task automation.

Managing an active autopilot

The Autopilot tab shows the current schedule, the next run time, and the total number of runs, along with these controls:
  • Run Now starts a run immediately without waiting for the next scheduled time and without changing the schedule
  • Pause and Resume stop and restart scheduled runs, keeping the configuration and run history intact
  • Edit changes the frequency, run time, or reporting mode of an existing schedule
Neo pauses autopilot automatically in two cases: when the account does not have enough credits to start a run, and after a run has failed on repeated consecutive attempts. The schedule and its history are preserved either way, so resolving the underlying problem and resuming picks it back up. Autopilot runs consume credits like any other task, and each run counts against the project spending cap if one is set. Start on a slower cadence and increase it once you can see what a typical run costs on your project.

Execution history

Each autopilot run appears under Execution History on the Autopilot tab, with the run time, status, and a short identifier. Runs show as pending, running, completed, or failed, with the run duration and any error message alongside. Click any entry to open the task that ran.

Opening and steering a run

An autopilot run is not a black box. Each run creates an ordinary task in the project, so it opens in the normal task view with the full plan, logs, insights, issues, files, and network activity for that run. You can reach it from Execution History, or from the project’s Tasks tab, where autopilot runs sit alongside the tasks you started yourself. Because these are ordinary tasks, they stay open after they finish. You can ask a follow-up question inside a completed autopilot run, push it further on something it found, or chain into a new task from where it stopped, exactly as you would with any task you started. That is also the fastest way to course correct. A follow-up you write inside an autopilot run belongs to that task, and completed tasks feed the context of later runs, so a correction you make there carries forward. Use project Settings for direction that should apply to every future run, and a follow-up inside a run for something specific to what that run just did.

Autopilot vs. task automation

Both features run Neo tasks automatically, but they serve different purposes. Use task automation when you have a specific, repeatable job to run on a schedule: a daily DAST scan, weekly recon sweep, or recurring API audit. You write the prompt and control what happens each time. Use project autopilot when the entire project is your unit of work and you want Neo to pick up where it left off. This is most effective when all tasks in the project share a common goal and the accumulated context makes each new run more informed than the last. For example: if your project goal is to test a web application, autopilot will use everything Neo has already discovered (endpoints, findings, completed scans, settings) to decide what to investigate next. It does not re-run prior tasks; it creates new ones.

Common use cases

Daily DAST scan — Create a task automation with your DAST prompt and set it to run daily. This gives you a fixed, repeatable scan with a prompt you control. Do not use autopilot for this. Ongoing security research project — Set up a project scoped to a target, add context in Settings (scope, objectives, known findings), run a few initial tasks, then enable autopilot. Neo will continue the research on a schedule, building on each previous run. Continuous penetration testing — Use autopilot on a project configured for a specific application. As Neo completes tasks and files findings, future autopilot runs incorporate that history and shift focus to unexplored areas.

Limitations

  • One autopilot per project. A project supports a single autopilot configuration; you cannot run multiple concurrently.
  • No prompt customization. The autopilot prompt is fixed. Use task automation if you need a specific prompt.
  • Project scope only. Autopilot is not available outside of a project.
  • Recurring schedules only. Autopilot cannot be set to run once at a future time; use a one time task automation for that.
  • Run times are UTC. The schedule does not follow your local time zone or adjust for daylight saving changes.