What is an Autopilot — and how is it different from a Subagent?
TenderB has two types of agents, and understanding the difference is key to setting up your workspace well.
Autopilot — runs complex, multi-step tasks autonomously. It operates at the orchestrating agent level: it plans, makes decisions, and can invoke Subagents to complete parts of the work. Think of it as your autonomous project driver.
Subagent — performs a single, focused task. It can be invoked directly by a user, or automatically by an Autopilot that is orchestrating a larger workflow.
From now on, TenderB supports multiple Autopilots in one workspace or team, so you can set up a dedicated Autopilot for each type of task — for example one for Go/No-Go analysis, one for proposal writing, and one for contract review.
How to create an Autopilot
Go to the Agent / Prompt management overview (accessible from the main navigation).
Click Create agent.
Under agent type, select Autopilot.
Give it a clear, recognisable name — for example "Proposal Writer – Offshore" or "Go/No-Go Analyser". A good name makes it easy for your team to pick the right Autopilot from the dropdown.
Assign it to a Category. Best practice: create a dedicated category for your Autopilot agents (e.g. "Autopilots") so they are easy to find and distinguish from Subagents in the overview.
Choose the Autopilot model (e.g. Marlowe Long for complex, long-running tasks).
Write the Goal / Prompt — describe what the Autopilot should accomplish (up to 25,000 characters).
Optionally: check Available to other users to share it with your team, or Add this agent to new teams by default to roll it out automatically.
Click Create agent to save.
Where to access your Autopilots
Once created, your Autopilots are available in three places:
Project creation flow — in the Autopilot step, open the agent dropdown and select the Autopilot you want to run for this project.
Autopilot button — click the Autopilot button and choose from the dropdown to start an autonomous run on the current project.
Chat interface — click Expand agents and select your Autopilot from the list to run it within an existing chat context.
Tips for a well-organised setup
Use clear, descriptive names — your team will see these names in every dropdown, so make them self-explanatory.
Group all Autopilot agents under a single Category (e.g. "Autopilots") in the management overview, so they are visually separated from Subagents.
Use Subagents for narrowly scoped, reusable tasks (extracting a deadline, scoring a criterion) and Autopilots for end-to-end workflows that combine multiple of those steps.