Factory AI Review 2026: Droids, Pricing & Limits

Disclosure: CodeToolPilot may earn a commission if you buy through links on this page, at no extra cost to you. How we make money.

Last verified: October 13, 2026. This is an editorial research review — how we review. Prices and plans change often; confirm on the vendor’s site.

Factory is a company that builds “Droids”, AI agents for software development that write code, review pull requests, generate tests, investigate incidents and keep documentation current. It calls itself an “agent-native software development platform”. The same agent runs in a desktop app, a terminal CLI, a web and mobile app, Slack, Microsoft Teams, Jira, Linear and CI pipelines. Its website now lives at factory.com, and the old factory.ai address redirects there.

Factory sells to individual developers, but much of its product is aimed at engineering organizations that want agents running across the whole delivery process, with model choice, admin controls and deployment options up to fully airgapped networks.

This is an editorial research review based on Factory’s pricing page, documentation, changelog, news posts and legal terms. We did not run our own benchmarks or timed tests.

Quick verdict

In our assessment, Factory is one of the broadest agent platforms available. The same Droid works from your terminal, a desktop app, chat tools, issue trackers and CI. It is model-agnostic, with Claude, GPT, Gemini and open-weight models plus bring-your-own-key support. Its enterprise controls and deployment options are unusually extensive. The trade-offs are pricing that is easy to start with but harder to predict, no free plan, and features such as Missions that work best on repositories already set up for agents.

Pro: One agent across the desktop app, CLI, SDKs, web, Slack, Teams, Jira, Linear and GitHub or GitLab CI.
Pro: Model choice: Anthropic, OpenAI, Google and open-weight models, an automatic Factory Router, and BYOK for your own keys or local models.
Pro: Strong governance: SSO, SCIM, zero data retention, audit logs, permission rules, sandboxing and airgapped deployment on higher tiers.
Pro: Built-in workflows beyond coding, including automated PR review, security review, AutoWiki documentation, incident response and scheduled automations.
Con: No free plan. Paid tiers use rolling rate limits and model multipliers rather than simple request counts.
Con: Business and Enterprise prices are not published, and SSO starts at Business.
Con: Missions need Extra Usage turned on, and Factory recommends a high “Agent Readiness” level for best results.
Con: The JetBrains integration is a chat front end that does not yet share open files, selections or diagnostics.

Visit Factory

What is Factory?

Factory’s core product is Droid, a coding agent that reads and edits files and runs commands where your code lives: a laptop, a CI runner, a VM or a Factory-hosted “Droid Computer”. You can drive it interactively in the Factory App (macOS and Windows) or the Droid CLI, run it headless with droid exec, embed it with Python and TypeScript SDKs, or delegate work to it from Slack, Microsoft Teams, Linear and Jira. Editor support covers VS Code, Cursor and Windsurf through an extension, and JetBrains IDEs and Zed through the Agent Client Protocol.

Around that agent, Factory has built what it calls a “Software Factory”: automations for triage, code review, QA, documentation and incident response. The company ships often: its changelog lists CLI v0.240.0 and Desktop v0.197.0 on October 13, 2026, and Automations reached general availability on September 30, 2026. Factory says it raised $200M at a $5B valuation in September 2026.

Key features

Droid across every surface

Factory says all products, including IDE, web, CLI and Slack, are included in every plan. Sessions can sync to the cloud, so you can check a task from a browser or phone. Droid Computers are persistent Factory-managed machines that keep packages, files and services between sessions.

Missions for large projects

Missions are Factory’s orchestration feature for multi-feature work. You plan features and milestones with Droid, then “Mission Control” coordinates worker agents, validates progress and lets you step in. The docs suggest the sweet spot is roughly 1 to 500 features. They also warn that repositories should be at Agent Readiness Level 4 or higher, so that the mission can test its own work.

Model independence and Factory Router

Factory offers current Claude, GPT and Gemini models plus open-weight options, each with a usage multiplier. Factory Router picks a model per task. Factory reports 63% aggregate cost savings versus frontier-model rates in its own evaluations, which we have not verified. BYOK lets you add your own OpenAI, Anthropic, Bedrock or OpenAI-compatible endpoints, including local models. Your keys stay in a local settings file, and custom models work in the CLI and desktop app but not on the hosted web or mobile apps.

Automated code review and security review

The /install-code-review command sets up a GitHub Action or GitLab CI component that reviews pull requests, posts inline comments and approves when it finds no issues. The /review command reviews local changes before you push. A separate security review workflow uses STRIDE and OWASP-style methods, and Droid Shield scans Droid’s own commits and pushes for secrets.

Safety controls

Autonomy levels (Off, Low, Medium, High) control how much Droid does without asking. Permission rules can allow, require approval for or block shell commands, and an optional OS-level sandbox limits file and network access. Hooks, skills, subagents, MCP servers and AGENTS.md files customize behavior.

Enterprise deployment and analytics

Factory documents cloud-managed, hybrid and fully airgapped deployment patterns, plus self-managed GitHub Enterprise Server and GitLab. Factory Analytics, Agent Effectiveness dashboards and OpenTelemetry export track adoption, cost and output.

Who should use Factory

Factory fits engineering organizations that want one agent platform for coding, review, documentation and operations, especially where model choice, governance or on-premise deployment matter. It also suits heavy individual users who want a multi-model agent across terminal, desktop and phone, and who are comfortable managing usage limits.

Who should skip it

Skip Factory if you want a free tier, a simple flat price, or just an in-editor assistant for autocomplete and quick edits. Small teams that need SSO will have to talk to sales.

Strengths and limitations

Strengths

Breadth. Few agents cover coding, headless CI, chat-tool delegation, PR review and documentation in one product.
No model lock-in. Native multi-vendor models, a router, open-weight fallbacks and BYOK reduce dependence on any one AI lab.
Enterprise depth. Airgapped deployment, hierarchical managed settings, audit logs and telemetry export suit regulated industries. Factory’s security page lists SOC 2, ISO 42001, GDPR and CCPA and says it does not use customer code as training data.

Limitations

Hard-to-predict costs. Rate-limit windows, model multipliers, Droid Core, Extra Usage and BYOK allowances make monthly spend harder to forecast than a flat seat price.
No free tier. The pricing page lists no free plan, and onboarding now asks for a payment method at signup.
Readiness matters. Factory itself says Missions need repositories with scriptable end-to-end testing. Many legacy codebases will need work first.
Uneven editor depth. VS Code-family editors get full context sharing. JetBrains currently gets a chat front end only.
Restrictive terms for evaluators. Factory’s terms bar customers from using the service for benchmarking or publishing performance information, so independent comparisons are scarce.

Pricing (as of October 2026)

Plan	Price	What you get
Pro	$20/month	Desktop app, CLI and SDK, cloud and local background agents, usage tracking and the agent-readiness dashboard, with baseline rolling rate limits.
Plus	$100/month	Everything in Pro, roughly 5x Pro usage and expanded rate limits. The pricing page lists Droid Computers here.
Max	$200/month	Everything in Plus, roughly 10x Pro usage and early access to new features.
Teams	$60/month per team + $40/seat	Pro limits for each seat, up to 10 seats, centralized month-to-month billing.
Business	Contact sales	Custom usage limits, SSO, SAML/SCIM, zero data retention, audit logs, basic admin controls and dedicated onboarding.
Enterprise	Contact sales	Unlimited members, dedicated compute, on-premise options, sub-organizations, data residency and priority SLAs.

Individual plans run on three rolling rate-limit windows (5-hour, 7-day and 30-day). When you run out, you can switch to the free “Droid Core” pool of open-weight models or buy prepaid Extra Usage, with a $10 minimum and no expiry. BYOK usage is free up to an allowance, then charged according to your plan. Factory’s documentation and pricing page differ slightly on which tier includes Droid Computers, so confirm current details on Factory’s pricing page. See Factory pricing

Prices verified on October 13, 2026 against the official Factory pricing page. Plans change often — confirm on the vendor’s page before you buy.

Use cases

Factory is built for delegating work to agents across many tools. These scenarios are based on Factory’s documentation and the research behind our review, not on hands-on testing.

Delegating tickets from Slack, Linear or Jira

A product team wants routine bugs and small features picked up without anyone opening an editor. Factory’s Delegations let you hand issues to Droid from Slack, Microsoft Teams, Linear or Jira, using shared instructions and compute. A developer should still review the resulting diff before merging.

Adding AI review to every pull request

A team on GitHub or GitLab wants an automated first-pass review. Running /install-code-review in Droid sets up a workflow that posts inline comments and approves clean pull requests. In our assessment, this is a reasonable fit for teams already paying for Factory. Teams that only want PR review may prefer a dedicated reviewer.

Large multi-feature projects with Missions

A tech lead plans a migration or a new module with many parts. Missions break the work into features and milestones, then coordinate worker agents and validate progress. The caveat is that Factory recommends a repository with scriptable end-to-end testing, and Missions draw on Extra Usage.

Agents inside a regulated or airgapped network

A bank or public-sector team cannot send code to a public SaaS. Factory documents hybrid and fully airgapped deployments using your own model endpoints, with managed settings, audit logs and telemetry export. These options sit on the Business and Enterprise tiers, which require a sales contract.

Alternatives

Claude Code: Anthropic’s agent across terminal, IDE, desktop and cloud, best if you are committed to Claude models.
OpenAI Codex: an open-source CLI, IDE extension and cloud tasks included with ChatGPT plans.
Amp: a multi-model agent with parallel cloud runs and the option to reuse existing subscriptions.
Cursor: a VS Code-style editor with cloud agents and Bugbot reviews, if you prefer an IDE-first workflow.

Final verdict

Factory is a serious, fast-moving agent platform that stands out for breadth, model independence and enterprise deployment options. Individuals can start at $20 a month, but the real value shows up when a team connects Droid to CI, chat and issue trackers. Expect to learn its usage model and to invest in agent readiness before the most autonomous features pay off.

Try Factory

FAQ

Is Factory free?

No. Factory’s pricing page lists no free plan. Individual plans start with Pro, and Teams adds a per-team fee plus a per-seat fee. When you hit your rate limit, you can keep working on the free Droid Core pool of open-weight models, but you still need a paid plan to use it.

Can I use my own API keys or local models with Factory?

Yes. Factory’s BYOK feature lets you add OpenAI, Anthropic, Amazon Bedrock or any OpenAI-compatible endpoint, including local models. Keys stay in a local settings file. Custom models work in the CLI and desktop app but not on the hosted web or mobile apps. BYOK usage is free up to an allowance, then charged according to your plan.

Does Factory work with JetBrains and VS Code?

Yes. VS Code, Cursor and Windsurf get an extension that shares your active file, selection and diagnostics. JetBrains IDEs and Zed connect through the Agent Client Protocol. Factory notes that the JetBrains integration does not yet share open files, selections or diagnostics automatically.

Is my code used to train AI models?

Factory’s security page says it does not use customer code as training data. Business and Enterprise plans add zero data retention, and enterprises can route model traffic through their own providers or gateways. If you bring your own keys, the model provider’s own terms also apply.

Factory vs Claude Code: which should I pick?

Claude Code is Anthropic’s agent and is built around Claude models. Factory runs Claude, GPT, Gemini and open-weight models and adds chat-tool delegation, automations and airgapped deployment. In our assessment, Factory suits teams that want model choice and a broader workflow platform, while Claude Code suits developers committed to Claude.

Factory Review (2026): Droid Agents From Terminal to CI