Evaluation record · factory-droids

Factory Droids

vDroids platform (2026, Missions era)

Factory AI

Agentautonomousenterprisesdlc-platformmulti-agent

Adequate

About This Agent

Factory AI's enterprise agent-native software development platform. Specialist autonomous Droids handle coding, testing, code review, refactoring, and DevOps work across Desktop, CLI, SDK, Slack, ticketing, and CI, with Missions enabling long-horizon multi-agent workflows. #1 on Terminal-Bench (Sept 2025); $220M raised at a $1.5B valuation (Apr 2026); used daily by hundreds of thousands of developers at enterprises including Nvidia, Adobe, and EY.

Last Evaluated: July 9, 2026

Official Website

Trust Vector Analysis

Dimension Breakdown

🚀Performance & Reliability

task completion accuracy

Review of public Terminal-Bench leaderboard placement (product-harness category) and sustained benchmark results; vendor-run submissions weighted with independent leaderboard verification

Evidence

Factory - Droid #1 on Terminal-Bench — Droid reached #1 on Terminal-Bench at 58.8%, outperforming Claude Code and Codex CLI; Factory harnesses held three of the top five slots (Opus 4.1 58.8%, GPT-5 52.5%, Sonnet 4 50.5%), showing harness design drives results across models

Factory leaderboards documentation — Factory maintains published leaderboard results and claims #1 status across leading software development agent benchmarks

highVerified: 2026-07-09

tool use reliability

Assessment of multi-surface tool execution (shell, editor, integrations) and MCP/tool configuration reliability

Evidence

Factory Series C announcement — Droids operate across terminal, editor, Slack, ticketing, and CI with full system access and local context via Factory Desktop; works with any model and any interface

mediumVerified: 2026-07-09

multi step planning

Evaluation of Missions long-horizon planning and spec-driven execution across multi-stage engineering tasks

Evidence

Factory Series C announcement — Missions capability enables long-horizon, multi-step, multi-agent workflows spanning stages of the SDLC

mediumVerified: 2026-07-09

memory persistence

Review of org context engineering, project guidance files, and cross-session context persistence

Evidence

Factory documentation — Organization-level context: AGENTS.md project guidance, agent-readiness dashboard, and persistent org/repo context shared across Droids and sessions

mediumVerified: 2026-07-09

error recovery

Inference from benchmark categories requiring iterative failure recovery plus documented test-iteration behavior

Evidence

Factory - Droid #1 on Terminal-Bench — Terminal-Bench tasks include debugging, legacy modernization, and build/test loops where Droid's harness recovered from failures better than competing agents even on sub-frontier models

mediumVerified: 2026-07-09

agent collaboration

Review of specialist-Droid division of labor, Missions multi-agent orchestration, and delegation surfaces

Evidence

Factory Series C announcement — Specialist Droids (coding, review, testing, docs, DevOps-style tasks) coordinate in multi-agent Missions; work is delegated from Slack, tickets, or CI and returns as reviewable output

mediumVerified: 2026-07-09

🛡️Security

tool sandboxing

Architecture review of local execution scoping, beta OS-level sandbox, and managed cloud sandbox model

Evidence

Factory CLI security documentation — Shell commands and file edits execute locally with only necessary context/diffs sent to Factory's cloud; optional OS-level sandbox with kernel-enforced filesystem/network isolation (Beta); Droid Computers provide Factory-managed cloud sandboxes on Plus/Max

mediumVerified: 2026-07-09

access control

Review of identity, RBAC, tool-permission granularity, and centrally managed policy controls

Evidence

Factory CLI security documentation — SAML 2.0/OIDC SSO with SCIM provisioning, role-based access controls, per-tool allow/ask/reject permissions, write access restricted to project directories, and enterprise-managed security policies

highVerified: 2026-07-09

prompt injection defense

Review of documented injection mitigations (rare among peers) tempered by absence of independent adversarial testing

Evidence

Factory CLI security documentation — Prompt injection detection and input sanitization documented as built-in, with risky operations requiring explicit user approval; effectiveness not yet validated by third-party research

mediumVerified: 2026-07-09

data isolation

Review of encryption, key management, ZDR, and tenant partitioning claims

Evidence

Factory CLI security documentation — TLS 1.3 in transit, AES-256 at rest with AWS KMS, customer-managed encryption keys (BYOK), zero data retention mode, and dedicated compute with partitioned inference on Enterprise

mediumVerified: 2026-07-09

open source transparency

Source availability assessment

Evidence

Factory documentation — Fully proprietary platform and harness; no source code published, though docs, leaderboard methodology, and a public trust center (trust.factory.ai) provide partial transparency

highVerified: 2026-07-09

🔒Privacy & Compliance

data retention

Review of no-training commitment and ZDR tier availability

Evidence

Factory CLI security documentation — Explicit commitment that Factory never trains on customer code; Zero Data Retention mode included on Business tier and above

mediumVerified: 2026-07-09

gdpr compliance

Compliance certification review (SOC 2 Type II, GDPR) via security docs and trust center

Evidence

Factory CLI security documentation — SOC 2 Type II certified with regular penetration testing and GDPR-compliant operations; compliance documentation via trust.factory.ai

mediumVerified: 2026-07-09

third party data sharing

Data flow analysis of multi-provider model routing and enterprise inference isolation options

Evidence

Factory pricing — Workloads route to third-party frontier models (GPT-5, Claude Opus/Sonnet, Gemini); Enterprise offers dedicated compute with partitioned inference to tighten the data path

mediumVerified: 2026-07-09

local deployment option

Deployment options assessment: on-prem/private-cloud available at Enterprise, no self-contained air-gapped offering

Evidence

Factory pricing — Enterprise tier offers on-premise deployment, data residency options, and customer-managed encryption keys; CLI executes locally by default, but the platform and inference are cloud services

mediumVerified: 2026-07-09

👁️Trust & Transparency

documentation quality

Documentation completeness and accuracy review

Evidence

Factory documentation — Thorough docs spanning CLI, security, enterprise administration, leaderboard methodology, and integration setup

highVerified: 2026-07-09

execution traceability

Review of session logs, OTel telemetry, and audit features for governance

Evidence

Factory CLI security documentation — Complete session logging maintained with OpenTelemetry metrics integration and enterprise audit-oriented administration

mediumVerified: 2026-07-09

decision explainability

Assessment of plan/diff visibility and change justification in delegated workflows

Evidence

Factory Series C announcement — Droids produce reviewable plans, diffs, and PR-style outputs; Missions expose multi-step progress, though internal harness decisions are less visible than open competitors

mediumVerified: 2026-07-09

open source code

Open source assessment

Evidence

Factory documentation — Closed-source product; no public source repositories for the platform or Droid harness

highVerified: 2026-07-09

community activity

Community and ecosystem engagement analysis

Evidence

Factory Series C announcement — Hundreds of thousands of daily developers and revenue doubling month-over-month for six months; community is enterprise-customer-driven rather than open-source contributor-driven

mediumVerified: 2026-07-09

⚙️Operational Excellence

ease of integration

Integration surface assessment across developer and team-workflow entry points

Evidence

Factory Series C announcement — Droids accessible from Desktop, CLI, and SDK, plus Slack, ticketing systems, and CI; works with any model and existing developer tooling

highVerified: 2026-07-09

scalability

Assessment of parallel cloud execution, enterprise seat scaling, and named large-scale deployments

Evidence

Factory pricing — Droid Computers provide Factory-managed cloud sandboxes for parallel background agents; Enterprise supports unlimited members with dedicated compute; proven at Nvidia, Adobe, EY, Palo Alto Networks, Adyen scale

mediumVerified: 2026-07-09

cost predictability

Pricing model analysis: flat tiers with opaque underlying usage allowances; enterprise seat+usage blend requires negotiation

Evidence

Factory pricing — Individual tiers Pro $20/mo, Plus $100/mo (~5x usage), Max $200/mo (~10x usage); Business/Enterprise custom-priced with tailored usage limits per seat. Usage-based consumption within tiers varies with task size

mediumVerified: 2026-07-09

monitoring capabilities

Review of usage dashboards, telemetry export, and admin governance tooling

Evidence

Factory pricing — Billing tracking and agent-readiness dashboards on all paid tiers; session logging, OpenTelemetry metrics, and enterprise admin controls for governance

mediumVerified: 2026-07-09

production readiness

Vendor viability and deployment maturity assessment from funding, revenue trajectory, and named enterprise adoption

Evidence

Factory Series C announcement — $150M Series C led by Khosla Ventures (Sequoia, Blackstone, Insight, NEA participating) at $1.5B valuation; $220M raised in total; revenue doubling monthly for six months with major enterprise customers

mediumVerified: 2026-07-09

Strengths

+#1 on Terminal-Bench (58.8%, Sept 2025) with three of the top five harness slots, beating Claude Code and Codex CLI
+Specialist Droids plus Missions for long-horizon, multi-step, multi-agent SDLC workflows
+Strong enterprise security package: SOC 2 Type II, SSO/SCIM, RBAC, allow/ask/reject tool permissions, ZDR, BYOK keys, on-prem
+Documented prompt injection detection and input sanitization - rare among coding agents
+Observability for governance: complete session logs, OpenTelemetry metrics, admin dashboards
+Proven enterprise adoption: Nvidia, Adobe, EY, Palo Alto Networks, Adyen; hundreds of thousands of daily developers
+Strong vendor trajectory: $220M raised, $1.5B valuation (Apr 2026), revenue doubling monthly for six months

Limitations

!Fully closed-source platform and harness; transparency relies on vendor documentation
!OS-level sandbox is still Beta; default local execution relies on permission gates
!Usage allowances behind flat tiers are opaque, making heavy-use costs hard to forecast
!Injection defenses and security claims lack independent third-party validation
!Workloads depend on third-party frontier model providers (GPT-5, Claude, Gemini)
!Rapid growth-stage company: pricing, packaging, and product surface have churned repeatedly since 2025
!Best experience requires adopting Factory's platform surfaces; deep vendor coupling for SDLC-wide automation

Metadata

license: Proprietary

supported models

0: OpenAI GPT-5

1: Anthropic Claude Opus and Sonnet

2: Google Gemini

3: Additional models via harness (model-agnostic design)

programming languages

0: Most major languages (Python, TypeScript, Java, Go, C++, etc.)

deployment type: Cloud platform + local CLI/Desktop execution; Droid Computers managed cloud sandboxes; on-premise available at Enterprise

tool support

0: Droid CLI, Desktop, and SDK

1: Slack and ticketing (Jira/Linear-style) delegation

2: CI/CD integration

3: MCP and custom tools

4: OpenTelemetry metrics

first release: Founded 2023; Droid CLI GA and Terminal-Bench #1 September 2025; Missions/Series C April 2026

pricing: Individual: Pro $20/mo, Plus $100/mo, Max $200/mo (usage-tiered); Business (up to 150 seats, ZDR, SSO) and Enterprise (unlimited, dedicated compute, on-prem) custom seat+usage pricing

company milestones: $220M total raised; $150M Series C at $1.5B valuation led by Khosla Ventures (2026-04-16); revenue doubled month-over-month for six consecutive months; customers include Nvidia, Adobe, EY, Palo Alto Networks, Adyen

Use Case Ratings

code generation

Terminal-Bench leader; specialist Droids cover coding, review, testing, and refactoring with strong results even on sub-frontier models

data analysis

Can script analyses and ML workflows in its environments, but the platform targets the software delivery lifecycle

research assistant

Codebase and ticket-context research is strong; general-purpose research is not the product's aim

customer service

Not designed for customer-facing conversational work; Slack presence is for engineering delegation only

Similar Agents

Devin

Cognition

Claude Code

Anthropic

OpenAI Codex

OpenAI

GitHub Copilot Coding Agent

GitHub (Microsoft)

Cursor Agent

Anysphere (SpaceX acquisition announced 2026-06-16, expected to close Q3 2026)