Factory builds autonomous "Droids" that plan, code, test, and ship engineering work, with enterprise traction on hard legacy-migration problems, and its own Factory Analytics and Signals now give cost, output, and session-quality visibility native to that runtime. Tetriz isn't a competing agent, and its own measurement doesn't reach into Factory's runtime either: across the six coding agents it does support, AI Drivers attributes every autonomous loop to an owner, with runs, cost, and outcome, and its Harness Control Plane publishes an approved AI configuration to Claude Code, Cursor, Codex, and Kiro, tracking each one's install status.
Factory's Droids are a capable autonomous execution layer, specialised agents that plan, code, test, and ship work across the software lifecycle, with enterprise traction on legacy systems most tools won't touch. Factory Analytics (cost and output mapped to Jira/Linear work) and Signals (an LLM-judge system that analyzes sessions at scale for friction and auto-files tickets) now give real, shipped measurement native to the Droid runtime. Tetriz doesn't build a competing Droid, and its own measurement doesn't reach into that runtime either. Where it operates instead is the six coding agents Droids don't touch: AI Drivers attributes every autonomous loop in Cursor, Claude Code, GitHub Copilot, Codex, Kiro, and Antigravity to an owner, with runs, cost, and outcome, and the Harness Control Plane publishes and tracks an approved AI configuration for Claude Code, Cursor, Codex, and Kiro.
| Capability | Tetriz | Factory |
|---|---|---|
| Loop & cost accounting | AI Drivers attributes every autonomous loop run in its six supported agents to an owner, with its runs, cost, and outcome; AI ROI turns that into an org-level net dollar figure, net of the AI Tax. | Factory Analytics maps cost and output to Jira/Linear work, a real cost-versus-outcome view, but it's scoped to Droid's own runtime, not the six agents Tetriz covers. |
| Session quality scoring | Six coachable dimensions (clarity, specificity, context, actionability, completeness, efficiency), scored from the session directly. | Signals uses LLM judges to detect friction and dead-ends in Droid sessions at scale, a real capability, and routes what it finds by auto-filing tickets rather than publishing a structured per-session score. |
| Weekly coaching loop | A weekly Session Quality Score returned directly to the engineer who ran the session, visible only to them and their org's admins, never a cross-engineer ranking. | Signals routes what it finds to a ticket queue rather than publishing a weekly signal to the engineer who directed the session. |
| Harness improvement & enforcement | A recommendation engine curates a vetted, pre-filtered base of skills, sub-agents, and rules; the Harness Control Plane publishes that baseline across the four coding agents it enforces on (Claude Code, Cursor, Codex, and Kiro) and tracks each engineer's install status against it. | Governs execution inside its own Droid runtime, which sits outside those six agents; it doesn't publish a harness baseline or track per-engineer install status there. |
Droids plan, code, test, and ship work end to end on hard problems, including disclosed legacy-migration work at MongoDB and Ernst & Young. Tetriz has no equivalent; it prices and improves the loop, it doesn't write the code itself.
Factory Analytics (cost, output, and per-user breakdowns mapped to Jira and Linear) and Signals (LLM-judge session analysis that auto-files tickets on friction) are real, shipped capabilities, scoped to Factory's own Droid runtime.
Not as an execution engine, and not as an analytics layer for Droid's own runtime either, Factory Analytics and Signals now cover that natively. Where Tetriz competes: across the six coding agents it supports, it prices what an autonomous loop cost against what it shipped, publishes and tracks the AI configuration those agents run with, and coaches the engineer directing the work every week.
No, not while the task runs inside Factory's own environment, that's what Factory Analytics and Signals are for. What Tetriz does read, cost and outcome included, is any session or autonomous loop run directly in one of its six supported agents: Cursor, Claude Code, GitHub Copilot, Codex, Kiro, or Antigravity.