Rating certificate · AW/26-09/49326 Sept 2026
60of 100
The Register / Coding agents / Factory Droid
Register entry · Coding agents · updated 26 Sept 2026
Factory Droid
Droid is Factory's coding agent. It runs as a terminal UI, a desktop app, a web app with cloud 'Droid Computers', and a headless `droid exec` mode for CI, and can use Anthropic, OpenAI, Google and open-weight models or bring-your-own-key endpoints. It adds spec mode, multi-agent Missions, autonomy levels, optional OS-level sandboxing and enterprise controls. Factory raised $200M at a $5B valuation in September 2026.
What reviewers and users praise
- Model-agnostic: users can mix providers and BYOK models
- High Terminal-Bench 2.0 placement for a product harness
- Very frequent releases and extensive documentation
What they criticise
- Usage-based plans can run out mid-Mission on Pro/Plus
- Coordinator/Mission overhead is heavy for small tasks
- Public issue tracker shows reliability bugs (silent no-op missions, MCP tools not propagated, BYOK image handling)
Experts vs the crowdneeds both scores
Experts71
Crowd—
| Price | From $20 |
|---|---|
| Use it via | CLI, Desktop app, Web app, IDE extension, API, Chat apps |
| Runs on | macos, linux, windows, web |
| Licence | proprietary |
| Updated | Evidence collected 26 Sept 2026 |
How the rating was worked out
Weights for Coding agents: adoption 35%, experts 40%, crowd 25%. Pillars without enough evidence are left out and the others re-weighted. The method.
| Adoption & momentumweight 35% | 48 | npm weekly downloads 8.4k → 25 · 61 releases in 90 days → 100reach × 0.7 + momentum × 0.3 |
|---|---|---|
| Expertsindependent reviews · weight 40% | 71 | 4 independent reviews (3 positive · 1 mixed · 0 negative), averaged toward 50 for small samples. Benchmarks, shown not scored: Terminal-Bench 2.0 No. 9 of 43 (Droid + GPT-5.3-Codex) · SWE-bench Verified No. 70 of 77 (Factory Code Droid + Undisclosed) · SWE-bench Lite No. 36 of 58 (Factory Code Droid + Undisclosed) · SWE-bench Test No. 9 of 13 (Factory Code Droid + Undisclosed) |
| Crowdcommunity sample · weight 25% | — | Community sample from a single platform |
| Rating | 60 | Provisional: low confidence (2 of 3 pillars, 6 community posts). |
Evidence checks. 4 GitHub issues not counted as sentiment. How evidence is checked.
Trust & safety
Seven dimensions graded A–D from documented facts. Overall: B.
B
Permission model
Documented autonomy levels (Off/Low/Medium/High) gate commands by risk, with allow/deny/block lists; `droid exec` is read-only by default. OS-level sandbox exists but applies 'once enabled' (opt-in). source
B
Data access scope
Runs in the user's repository with shell access; sandbox can restrict filesystem and network via allowedDomains and path policies when enabled. Broad by default, documented. source
B
Data storage
Cloud-managed mode stores operational logs/metrics at Factory; hybrid and fully airgapped deployments are documented for enterprises; retention/residency tunable per engagement. source
B
Data retention & training
Security page states Factory does not use customer code as training data; retention periods are deferred to the trust center, not stated in docs read. source
A
Incident history (24 months)
No CVEs or security advisories found (NVD keyword search, Factory-AI/factory repo) in the last 24 months. source
B
Compliance
Security page lists SOC 2, GDPR, ISO 42001 and CCPA and links a trust center (trust.factory.ai); SOC 2 report type could not be verified because the trust center did not render. source
B
Transparency
Closed source; very detailed dated public changelog and documentation. source
The experts
4 named independent reviews found; dated reviews from the last 12 months are scored.
- positive
Droid is worth trying if you spend meaningful time in the terminal, work across large or multi-service repos
Developers Digest · 10 Jun 2026 · developersdigest.tech - mixed
The platform excels in environments that match the coordinator-droid model and struggles where that model feels like overkill or misaligned.
Digital Applied · 13 Apr 2026 · digitalapplied.com - positive
Droid’s core thesis: agent architecture beats raw model selection when it comes to top-tier performance.
Murat Akdeniz (Codexpedite) · 6 Jul 2026 · codexpedite.com - positive
If your team is looking to automate feature implementation, code review, testing, and bug fixing, Factory can serve as a reliable assistant.
Brenda Stokes Barron (Fritz ai) · 22 Jan 2026 · fritz.ai
The crowd
6 coded posts from hackernews.
5 positive1 mixed0 negative
- mixed
For people seeking OS-native harnesses, I can recommend Factory's Droid.
Hacker News comment · 18 Sept 2026 · news.ycombinator.com - positive
Droid BYOK + GLM4.7 seems like a really cost effective backup in the little bit I have experimented with it.
Hacker News comment · 23 Dec 2025 · news.ycombinator.com - positive
Factory’s droid is pretty good for a cross-provider solution.
Hacker News comment · 12 Jan 2026 · news.ycombinator.com - positive
That's what I like with Factory.ai droid: making the spec with one agent and implementing it with another agent.
Hacker News comment · 16 Feb 2026 · news.ycombinator.com - positive
Both are great, and so is Factory Droid.
Hacker News comment · 3 Apr 2026 · news.ycombinator.com - positive
Same model, opus, works better in 3P harnesses such as Factory Droid or Amp.
Hacker News comment · 4 May 2026 · news.ycombinator.com
Pricing
As published, checked 26 Sept 2026.
| Plan | Price | Notes |
|---|---|---|
| Pro | $20 / monthly | Individual; desktop/CLI/SDK, cloud and local agents. |
| Plus | $100 / monthly | ~5x Pro usage; Droid Computers. |
| Max | $200 / monthly | ~10x Pro usage; early feature access. |
| Teams | $60 / monthly | $60/mo base + $40/mo per seat; up to 10 seats. |
| Business | Custom | Up to 150 seats; SSO, SAML/SCIM, audit logging. |
| Enterprise | Custom | Unlimited members, dedicated compute, on-premise deployment, data residency. |
Recent changes
- 26 Sept 2026
CLI v0.228.0: /migrate workflow command, command palette rebuild, in-place context compression. source
- 19 Sept 2026
CLI v0.223.0: session forking, PyCharm support, Databricks connector. source
- 15 Sept 2026
Raised $200M at a $5B valuation; names Nvidia, Blackstone, RBC, Palo Alto Networks, Adobe and T-Mobile as customers. source
- 7 Aug 2026
CLI v0.190.0: Claude Opus 5 in Auto model, HTTPS proxy support, GitLab CI setup. source
- 7 Jul 2026
CLI v0.167.0: port-forwarding for Droid Computers. source
Ranked alongside
Other ranked entries in Coding agents.
OpenCodeAnomaly75
Claude CodeAnthropic73
Zed (Agent Panel)Zed Industries73
GitHub CopilotGitHub (Microsoft)72
Codex (CLI and app)OpenAI71
Sources
Every figure above traces to one of these, observed up to 26 Sept 2026.
- Reuters (search snippet): AI coding agent startup Factory triples valuation to $5 billion
- WSJ (search snippet): Factory raised $200 million at a $5 billion valuation
- Apr 2026: $150M Series C at $1.5B valuation (Khosla, Sequoia) - search snippet
- Factory-AI/factory: public docs and issue tracker (30 stars), not source code
- npm @factory/cli (7,854 weekly downloads, same CLI; not used for reach)