NewLaunching v1.0 with $15M led by Insight Partners

Run Safe AI Software Factories

Infrastructure and tools for engineering teams to automate repeat manual development work without compromising data security or software quality.

For agents

$npx skills add islo-labs/skills

Code gen is easy. Trusting AI is hard
Islo Enterprise Software Factory gives you control over what models do, see, and access, and the gates they have to pass.

Deterministic workflows

Islo Factory Lines adhere to your process.
  • Define a real engineering process, like implement → verify → review → risk check, and agents follow it run after run.
  • Scripts, reviewer agents, evals and human gates decide exactly when AI-written work ships.
  • Every PR arrives with the tests, recordings and logs to prove it works. You still own the merge.

Isolated execution

Islo Machines contain and control AI work.
  • Persistent machines with your full stack: 10+ repos, services and dependencies, with sized CPU, RAM and storage.
  • Start from your own image and snapshots, so work begins with the stack already installed.
  • Built by Incredibuild, which has spent decades in the CI/CD critical path of enterprise engineering teams like Bosch, BBVA and Intel.

You’re in control

You choose the model, the harness, the cloud.
  • Pick Claude Code, Codex, Cursor or your own harness for each line.
  • Use your own API keys, or Islo inference powered by Databricks, across every major proprietary and open source model.
  • Run on Islo cloud or your own. Try a new model the day it ships, and see if it does better on your work.

Turn workflows into Factory Lines

Automate partially or totally, at the pace that fits your team with humans-in-the-loop where you choose.

Bug fixing

Agents pick up an issue, reproduce it, write new tests if needed, develop a fix that passes, then open a PR.

INIssue labeled for an agent
factory line
OUTReproduced, tested, fixed, PR opened
“The moment I could trust that what comes back actually works and my only job was merging the PR, that was a huge change for us.”
Matan Kleyman · CEO, Obert
Explore Bug fixing →

Choose which steps are deterministic

Each stage of a Factory Line can be executed by an agent where the work needs judgment, or by scripts, rules, and decisions where it needs to be exact.

PROMPT
An agent, for judgmentAn AI coding agent follows your prompt: implement a feature, review a diff, investigate an alert. Pick the harness and model per stage.implement · Codex · gpt-6.1
SCRIPT
A script, for certaintyDeterministic commands for tests, builds, linters and checks. Same input, same result, every run.$ npm test && npm run lint
DECISION
A decision, for controlA rule or a decision agent routes the line on results: score the PR’s risk, then merge, retry, send it back, or ask a human to approve.risk: low → merge · high → ask a human

Control what agents see and record everything

Islo keeps agents from becoming an attack vector or exfiltrating code, data, IP, credentials or secrets, and gives your security team the audit trail to prove it.

SOC 2
TYPE II
SOC 2 Type IIIndependently audited controls
Your cloud (BYOC)Code and secrets stay in your VPC
Part of IncredibuildTrusted by 400+ enterprises
Gateway credential injectionAgents only ever see placeholder keys. The gateway swaps in the real key on the way out, so raw secrets never enter the sandbox.
Scoped secretsChoose exactly which keys each sandbox can use. Non-sensitive config lives separately as environment variables.
Network and egress policiesBlock specific actions at the network level, such as deleting tickets or closing PRs, whatever the agent tries.
Hardware isolationEvery sandbox is an isolated microVM with no access to the host or other organizations, and shuts down when the work is done.
Full audit trailEvery session, action, outbound call and dollar of usage is logged and attached to the run.
Cost limitsSet spend limits per line, so faster AI adoption doesn’t blow through budgets.

Teams running Islo Enterprise Software Factories

All customer stories →
Alien

Alien ships its product into each customer’s own cloud, so every release has to work the first time. Its QA line spins up a fresh machine per example project, where a Codex agent installs the product and uses it like a real developer. Once QA was trusted, Alien extended the factory to issue triage and bug fixing.

“Islo proved for me that the dream of having an actual software factory is realistic.”
Alon Gubkin · CEO, AlienRead the full story →
Obert

A small AI-native team competing with companies about 100x its size. Agents work in a prepared environment with Obert’s full stack: they reproduce and fix customer-reported bugs, run tests and open a PR. Some fixes now ship in minutes.

“The moment I could trust that what comes back actually works and my only job was merging the PR, that was a huge change for us.”
Matan Kleyman · CEO, ObertRead the full story →

Build Safe Software Factories