Factory AI is a proprietary SaaS platform for autonomous software development that deploys LLM-agnostic "Droids" — agents that handle complete SDLC tasks including feature development, refactoring, large-scale migrations, incident response, and code review. GA since September 2025, $70M raised (Sequoia, NEA, NVIDIA, J.P. Morgan), and production deployments at MongoDB, Zapier, Bayer, EY, NVIDIA, and Clari. Ranks #1 on Terminal-Bench at 58.8%, outperforming Claude Code (43.2%) and Codex CLI on complex full-SDLC tasks.
Why It's in Trial
Factory occupies the most ambitious position in this part of the radar: not just issue-to-PR automation but full lifecycle agent tasks — migrations that take hours, incident response that spans multiple services, systematic refactors across large monorepos. The enterprise customer list ($70M Series B co-led by NEA and Sequoia, with NVIDIA and J.P. Morgan) and the Terminal-Bench #1 ranking are the strongest signals this category has produced.
Trial rather than Adopt because:
- GA since September 2025 — only ~7 months at time of review; production track record is still developing
- Benchmark nuance: Terminal-Bench #1 (58.8%) demonstrates broad SDLC capability, but SWE-bench Full is 19–22% — significantly below frontier models (Gemini 3.1 Pro 78.8%, Claude Opus 4.6 80.8%); Factory is optimised for workflow integration, not isolated coding benchmarks
- Vendor-reported metrics (31x faster delivery, 96% shorter migration times) require independent verification
- Pricing is opaque: token-based consumption billing, no public price list, requires sales contact
- Proprietary SaaS only — no open-source option, no self-hosted path
When Factory is the right choice: Enterprise teams with complex migration and refactoring work that doesn't fit the "one discrete ticket" model of Sweep or the "CI-mature, Linear-first" prerequisites of Symphony. Teams already using Jira, Linear, Notion, or GitHub Issues who want an agent that works across all of them. Teams who need model-agnostic agents (Droids run on GPT-5, Claude Opus 4.1, Gemini 2.5 Pro, or BYOK).
When to look elsewhere: Teams wanting open-source self-hosted tooling (use OpenAI Symphony or Open SWE); teams with tight budgets and simple discrete tasks (use Sweep AI); teams where SWE-bench coding performance is the primary metric (Claude Code, OpenHands).
What Droids Do
Factory's agents are called Droids, and they differ from most coding agents by covering the full SDLC rather than just code generation:
- Code Droid: Feature development and bug fixes — reads codebase, implements changes, submits PRs with full execution traces
- Migration Droid: Large-scale systematic refactors (e.g. framework migrations, API version upgrades across thousands of files); claimed 96% reduction in migration time
- Review Droid: Automated code review with task context, not just static analysis
- Incident Droid: On-call response — reads runbooks, diagnoses production issues, proposes or executes remediations; claimed 96% reduction in on-call resolution time
All Droids operate in isolated sandboxes with full audit trails, reversibility via version control, and explainable action logs. The model-agnostic architecture uses modular adapters per LLM, which is how Droid + Claude Sonnet 3.5 was reported to outperform direct Claude Code usage on the same tasks.
Key Characteristics
| Property | Value |
|---|---|
| Interface | IDE, CLI, Web, Desktop app (macOS/Windows), Slack, GitHub |
| Provider | Factory AI (San Francisco; founded 2023) |
| License | Proprietary SaaS |
| Pricing | Token-based consumption billing; no public price list (contact sales) |
| Underlying model | Model-agnostic: GPT-5, o3, Claude Opus 4.1, Claude Sonnet 4, Gemini 2.5 Pro, BYOK |
| Issue tracker | GitHub, Jira, Linear, Notion |
| Terminal-Bench | 58.8% (#1) — outperforms Claude Code (43.2%), Codex CLI (terminal-bench.ai) |
| SWE-bench Full | 19–22% (optimised for workflows, not isolated tasks) |
| Funding | $70M total — Series B (Sep 2025): NEA, Sequoia, NVIDIA, J.P. Morgan |
| Notable customers | MongoDB, Zapier, Bayer, EY, NVIDIA, Clari, Bilt Rewards |
| GitHub | Factory-AI (proprietary, closed repos) |
| Website | factory.ai |
| Docs | docs.factory.ai |