Factory AI
Software development agents called droids
About Factory AI
Factory AI builds software development agents it calls droids, and the framing is deliberate: a droid is something you assign work to, not something you chat with. The platform is agent-native rather than an assistant added to an existing editor, and it spans desktop, CLI and an SDK so the same agents can be driven from a UI, a terminal or your own code.
The distinction that matters commercially is where the agents run. Factory offers both cloud droids, which are managed machines you delegate to, and local background agents, which work on your own hardware. Most competitors pick one. Offering both addresses the two objections enterprises actually raise: latency and control on one side, and not wanting proprietary code on someone else's machine on the other.
Usage is measured in hourly droid allocations rather than tokens or requests, which is an honest fit for the workload. Agentic engineering work is bursty and long-running, and hours map to it better than a per-message count that penalises a thorough agent.
There is no published free tier, which tells you the intended buyer. This is sold to teams and companies rather than to individuals evaluating on a Sunday, and the pricing ladder from $20 to $200 with separate Teams, Business and Enterprise tiers reflects that.
You pick a surface. The desktop app is the richest, the CLI suits developers who prefer to stay in a terminal, and the SDK exists so droids can be invoked from your own tooling and pipelines rather than only by a human.
You assign work to a droid. Because the platform is built around delegation rather than completion, the unit is a task with a goal, not a cursor position waiting for a suggestion.
Droids execute in the cloud on Factory-managed machines, or locally as background agents when the code should not leave your environment. Cloud droid time is what your hourly allocation buys.
Higher tiers raise the allocation rather than unlocking different features, so the pricing question is throughput: how much agent time your team consumes in a month.
- •Droids as the unit of work - agents you assign tasks to, rather than an assistant you prompt line by line.
- •Cloud and local execution - managed cloud machines for delegation, local background agents when code cannot leave your environment. Few competitors offer both.
- •Desktop, CLI and SDK - three surfaces on one platform, with the SDK meaning droids can be called from your own automation rather than only by hand.
- •Hourly allocation model - usage measured in droid hours, which suits long-running agentic work better than per-message metering.
- •Team and enterprise tiers - Teams at a base plus per seat, with Business and Enterprise adding custom terms through sales.
- •Multi-platform, agent-native design - built around agents from the start rather than retrofitted onto an editor.
Engineering teams that have decided agentic development is worth budget are the target. The absence of a free tier is a signal, not an oversight: this expects you to have already concluded the category works and to be choosing a vendor.
Companies with code that cannot leave their environment are the clearest fit, because local background agents remove the objection that usually kills these tools in security review, while cloud droids remain available for work where it does not matter.
Platform teams building internal automation will care most about the SDK, since it lets droids be triggered by pipelines and internal tools rather than only by a person at a keyboard.
Individual developers exploring the category should probably start elsewhere. With no free tier and pricing that starts at $20 and climbs to $200, the cheap way to form an opinion about agentic coding is a tool with a free tier, then come here when you know what you need.
Pricing
$20 - $200/mo
- Pro$20/mo
- Teams$60/mo
- Plus$100/mo
- Max$200/mo
From the vendor pricing page, 2026-09-05
