Factory Droid

A coding CLI with risk-based approvals and optional headless Droid Exec runs for scripts and CI.

No product visual on file

Choose it when

You want an interactive coding agent with explicit command policy, plus a read-only-by-default task runner you can opt into edits and CI writes.

Skip it when

You need a proven always-on service with a native scheduler and failure owner, or assume every edit waits for approval and local execution keeps all context off model providers.

01 / Classification

Autonomy evidence

This is our classification, not a vendor badge. The record says what the product normally does. The gaps stay visible.

How we classify it

Tool use + multi-step plans, still human-gated per session.

Runs unattended

No — the current record treats it as session-gated.

Trigger

A user starts Droid in a repository and submits a task. Optional Exec accepts a prompt, file or pipe for a headless task; CI scheduling is external orchestration, not the interactive CLI's normal trigger.

State

Sessions can be resumed or forked by ID. Exec can use a separate Git worktree; dirty worktrees are preserved for review. That is session/history and filesystem state, not a claim of unlimited cross-session memory.

Tools

Reads repository files, proposes and applies code changes, and runs shell commands. Exec starts read-only; explicit low enables file edits, medium adds local development operations such as installs and commits, and high permits remote writes/deployments under remaining rules.

Human control

Review diffs and choose an approval risk level. The quickstart's approval loop isn't universal: Low or higher permits file edits without asking. Effective command blocks reject; asks remain at High. Runtime safeguards and optional sandbox prompts are separate; organization admins can cap autonomy. Exec's one-shot path can't collect interactive approval, so High doesn't clear an effective ask.

02 / Operations

Control reality

Useful agents eventually touch something real. This is the operational picture we have, not a promise hidden in a sales page.

Where it runs

CLI commands and file edits execute on the machine where Droid runs. Optional Exec fits caller-run scripts, CI runners and worktrees; this listing doesn't extend that scope to Factory-hosted products.

Data boundary

Prompts and selected context follow the configured model and gateway even though commands/edits execute locally. Local execution isn't local-only data or a retention guarantee. Optional filesystem/network sandboxing is separate from command approvals; review provider and plan terms before sensitive work.

Identity & permissions

Interactive setup uses browser sign-in to Factory; security docs describe OAuth. Exec/CI uses a Factory API key. Keep runner credentials least-privileged; account entitlement and configured endpoint access weren't tested.

What it can write

local project files · shell commands · local Git changes when enabled · remote writes and deployments at higher Exec permission tiers, subject to remaining rules

Surfaces

Interactive terminal · Droid Exec scripts and CI

Export / lock-in

Exec emits stdout/stderr and structured JSON, including session IDs, and can write requested artifacts. Git files/worktrees provide local change artifacts to inspect. This isn't a verified full-account export or portable restore guarantee.

03 / Receipts

Evidence ledger

Here is what this brief is built on. “Documented” means we found a source. “Tested” means this desk actually ran it. Those are different things.

Evidence status

documented

Source set

6 primary sources

Evidence review

2026-10-18

Sources on file

Corrections

Not recorded yet. If we change a material claim, the correction belongs here.

04 / Change log

Related coverage

The brief tells you what the product is. Payloads track the changes that may alter the decision.

No linked change coverage on file yet.