AI has completely changed how we build software, at a pace we have never seen before.

Surviving this era means moving at its pace. Our tools were built for a slower one.

IDE is dead

LonglivetheADE

Agentic Development Environment

factorai

Forged for the agentic era.

the app

Three agents, one afternoon.

Hand each task to an agent and keep every one in view. A dot turns amber the moment one needs you; switch to it, answer, and move on. The Changes panel shows what it touched, the graph shows what it landed, and every line of it is yours to audit, whenever you choose.

what it is

You supervise, decide, review, and set the rules. Agents do the rest.

fix flaky e2e on CImigrate settings to SQLitewrite release notes

Bash pnpm e2e --grep login✓ 5 passed

Edit src/store/prefsStore.ts? Keep a backup of settings.json? (y/n)

Read specs/roadmap/DONE.mdEdit CHANGELOG.md

sessions

Run several agents at once

One tab per agent, each a real claude CLI in a real terminal. Give every one a task, switch between them as they work in parallel, and get more done in a day than one pair of hands could.

nightly triageNightlyHourlyCustom

0 2 * * *Triage new issues, label them, draft replies

nextTomorrow 02:0002:00 · now
thenSat 02:00 · Sun 02:00 · Mon 02:00

nightly triage waiting for 02:00nightly triage running, no tabnightly triage done · 4 issues labelled, 2 replies drafted

routines

Agents on a schedule

A prompt with a cron behind it. It fires without taking a tab, and catches up what it missed while you were away.

auth middlewarethe permission prompt fires before
release 0.44permission prompt on first launch
sidebar raila permission prompt in the footer

Every message, every session

Full text across your whole transcript history, read straight from ~/.claude. Nothing imported, nothing copied.

Changes3Graph

src/session.rs+12 -3
src/graph.ts+41 -8
specs/05-features.md+6 -0

✓ mainfix: wait for the session cookie before asserting

feat: session tabs come back on launch

fix: graph lanes keep their colour

chore: bump tauri to 2.11

audit

Audit whenever you choose

Agents commit as they go; nothing waits on you. Changes shows what was touched, the graph shows what landed, and every line is traced, ready for you on your own schedule.

▾Pro

BAbilling-api+

DSdocs-site+

▾Side projects

FAfactorai+

HOhomelab+

workspace

Arranged the way you think

Drag projects into groups you name. Dozens of projects and hundreds of sessions is the normal case, not the edge.

▾src
session.rs
graph.ts
generated.ts
secrets.enc.yaml

secrets.enc.yaml

files

Read, edit, decrypt

A tree with git decorations, a Monaco viewer, Markdown and PDF rendered, SOPS files opened in place.

fix flaky e2e on CI

~/dev/billing-apisplit

$ cargo test --workspacerunning 128 teststest session::spawn … oktest graph::lanes … oktest result: ok. 128 passed

$ pnpm devVITE ready in 312 ms➜ http://localhost:1420/

shell

Your shell, beside the agent

Your own terminal in the project directory: a test loop, a dev server, a git log. It belongs to the project and outlives the session above it.

local

Local, private, open source

Runs on your machine and reads the CLI's own files. No account, no server, no telemetry. MIT licensed, and it stays that way.

get it

Forge your own.

Download for macOS

Universal .dmg, Apple Silicon and Intel · updates itself · needs the Claude Code CLI

about

factorai

factorai is built by engineers who spend the day supervising agents and wanted the tool arranged around that, not around a cursor.

It is open source, and it will stay open source: a powerful development tool should be available to everyone, not gated behind a seat.