Overview
Factory AI is an agent-native software development platform for engineering teams, centered on an autonomous coding agent called Droid. Droid can read codebases, plan tasks, modify files, run commands and tests, review changes, and carry the same context across the desktop app, Droid CLI, web sessions, remote computers, and editor terminals. It is officially positioned as a continuous loop covering signal intake, triage, planning, coding, verification, release, and monitoring, with outputs including working code, automated PR reviews, test results, documentation, deployment output, and incident reports. On September 30, 2026, Automations officially opened to all users, allowing users to describe workflows in natural language and have Droid execute them automatically on a schedule, Slack messages, GitHub events, or webhooks, with support for choosing your own model and your own machine. The platform offers four deployment forms: public cloud, hybrid, on-premises, and air-gapped.
Key Features
- Agent-Native Development Loop: Built around the Droid autonomous coding agent, it can read codebases, plan tasks, modify files, run commands, run tests, and review changes, covering a continuous loop from signal intake, triage, planning, coding, verification, release, to monitoring.
- Automations Natural Language Workflows: Officially opened to all users on September 30, 2026. Users describe workflows in natural language, and Droid can execute them automatically based on trigger conditions without manual supervision. An Automation consists of five parts: trigger, instructions, run identity, run target, and visibility.
- Four Types of Triggers: Supports scheduled triggers (natural language or five-field cron, executed in UTC), Slack messages (triggered by top-level messages in specified channels, with support for keyword inclusion and exclusion and channel name wildcard matching), GitHub events (PR opened, new comment, branch push, label change, check completed), and Webhooks (external services send an HTTP POST to a dedicated URL to trigger).
- Model Independence and Choose Your Own Machine: Not tied to a single model provider, with options including OpenAI, Anthropic, Google, xAI, and multiple open-source models, support for custom endpoints and bring-your-own API keys, and the ability to use Factory Router to select models automatically by task or by rules. Run targets can be Droid Computer, the desktop app on your own machine, or GitHub Actions.
- Multiple Run Surfaces and Editor Compatibility: Covers the Factory desktop app, Droid CLI, web, Droid Exec (headless mode running in CI/CD pipelines), and Droid Computers (Factory-hosted persistent cloud machines, also supporting bring-your-own hardware and local GPUs for running local models). On the editor side, it supports the VS Code family, JetBrains, and Zed, and also adapts to editors such as Vim in terminal scenarios.
- Security and Deployment Forms: Droid reads the live file system and does not need to store a static copy of the code in the cloud. Deployment forms cover Factory-hosted public cloud, hybrid, on-premises, and fully air-gapped (no external network, can run on AWS, Azure, or GCP). Compliance credentials listed include SOC 2 Type II, ISO 27001, and ISO 42001.
Use Cases
- Engineering teams use Automations to automatically move tickets forward to opening a PR, for example the Ticket to PR template picks up tickets from Linear or Jira every 15 minutes and advances them to opening a PR.
- Development teams use the PR Reviewer template to review new Pull/Merge Requests every half hour on weekdays and post reviews against the head commit, or use Automated Code Review to automatically post inline reviews for every PR.
- Ops and on-call teams use the Error Triage template to poll error sources such as Sentry and Datadog every hour and open fix PRs, or use Alert Responder and Incident Response to handle alerts and incidents.
- Teams use Daily Status Digest to push a to-do summary at 9 a.m. on weekdays, or use templates such as Slack Digest, Channel Q&A, and Slack Bug Fixer to complete summaries, Q&A, and fixes within Slack.
- Teams use the Code Health template to rotate through dependency audits, dead code, coverage gaps, and outdated documentation at 6 a.m. on weekdays, landing one maintenance PR each time; or use AutoWiki to automatically maintain repository documentation as code changes.
Pros
- The agent-native positioning is clear, with Droid covering a continuous loop from signal intake, triage, planning, coding, verification, release, to monitoring, and outputs including working code, automated PR reviews, test results, documentation, deployment output, and incident reports.
- Automations officially opened to all users on September 30, 2026, allowing workflows to be described in natural language and executed automatically on a schedule, Slack messages, GitHub events, or webhooks without manual supervision.
- Strong model independence, with options including OpenAI, Anthropic, Google, xAI, and multiple open-source models, support for custom endpoints and bring-your-own API keys, and the ability to use Factory Router to select models automatically by task or by rules.
- Broad run surface and editor compatibility, covering the desktop app, Droid CLI, web, Droid Exec, and Droid Computers, with editor support for the VS Code family, JetBrains, and Zed, and also adapting to editors such as Vim in terminal scenarios.
- Flexible deployment forms, covering Factory-hosted public cloud, hybrid, on-premises, and fully air-gapped, with compliance credentials listed including SOC 2 Type II, ISO 27001, and ISO 42001.
- Officially includes a set of out-of-the-box Starter templates covering scenarios such as Ticket to PR, PR Reviewer, Daily Status Digest, Code Health, Meeting Notes Sync, and Error Triage.
Pricing
Public pricing on the official website as of October 2026: Pro is $20 per month, including desktop/CLI/SDK, cloud and local background agents, usage statistics, and the Agent Readiness dashboard; Plus is $100 per month, with a rolling usage cap of about 5 times Pro, and access to Factory-hosted Droid Computers; Max is $200 per month, with usage about 10 times Pro, and early access to new features; Teams is $60 per month plus $40 per seat, with centralized billing for up to 10 seats; Business is a custom plan including custom usage limits, dedicated onboarding, SSO, SAML/SCIM, zero data retention, audit logs, and basic administrative controls; Enterprise adds unlimited seats, dedicated partitioned inference compute, on-premises deployment options, sub-organizations, data residency, and a dedicated account manager and SLA support. For pricing information not provided, refer to the official website.
Summary
Factory AI is an agent-native software development platform for engineering teams. Its core Droid can run across the desktop app, command line, web, IDE, and CI/CD, and carry the same context across multiple run surfaces. On September 30, 2026, Automations officially opened to all users, supporting four types of triggers: scheduled, Slack messages, GitHub events, and webhooks, as well as choosing your own model and your own machine. The platform offers four deployment forms: public cloud, hybrid, on-premises, and air-gapped. Model choice is not tied to a single provider, and pricing ranges from Pro at $20 per month to custom Enterprise plans.
Version History
- Droid 1.0 makes programmatic tool calling the default (2026-10-31): Factory shipped Droid 1.0, the largest update to its coding agent harness. Script mode is now on by default, letting the model write JavaScript that runs tools inside a sandbox, so one model request can drive hundreds of tool calls and fewer round trips are needed.
- OpenCompute launches: a self-hosted orchestrator for agent sandboxes, now in private preview (2026-10-14): On October 14, 2026, Factory introduced OpenCompute, a self-hosted orchestrator for agent sandboxes running on infrastructure the customer controls, now in private preview. It brings on-demand provisioning and lifecycle management into customer-controlled environments and integrates with Environments, Automations and Remote Control: an alert can trigger an investigation without someone provisioning a VM first, a burst of Slack delegations does not leave a team counting machines, and completed tasks release resources according to policy. OpenCompute runs inside the customer environment and establishes an outbound TLS WebSocket connection to Factory, proving possession of a registered private key by signing a fresh challenge; the key stays under customer control and can live in a supported cloud KMS, with no inbound firewall port required. Its Go control plane records desired and observed computer state in a Postgres database in the customer environment, and a reconciler continuously drives state together, so a control-plane restart does not erase unfinished work. Factory reports six weeks from prototype to first customer deployment, with drivers validated across AWS Lambda MicroVMs, Kubernetes, EC2, Agent Sandbox and Agent Substrate.
- Environments reaches general availability: reusable sandbox definitions, prebuilt images, rebuilt and validated every few hours (2026-10-12): On October 12, 2026, Factory announced that Environments is now generally available. An Environment is a reusable definition of the sandbox a Droid runs on, composed of the repositories, tools and dependencies it needs. Factory builds Environments into portable images ahead of time so each sandbox is ready before the Droid arrives, and rebuilds and validates them every few hours, promoting only successful builds. Combined with Automations, each run starts on a fresh sandbox from the last validated build, covering schedules, Slack messages, GitHub events and webhooks. The accompanying Agent Readiness framework scores each repository out of five, where five means a repository is fully ready for agents to be maximally productive; Factory notes most enterprise customers start at level two.
- Automations general availability (2026-09-30): Factory opened Automations to all users: describe a workflow in natural language and Droid runs it automatically on a schedule, a Slack message, a GitHub event, or a webhook, with a choice of model (including BYOK and Factory Router) and a run target (Droid Computer, your own machine, or GitHub Actions). Each automation consists of a trigger, instructions, run identity, run target, and visibility, and ships with ready-made starters such as Ticket to PR, PR Reviewer, Error Triage, and Docs Refresh