One Orchestrator, Many Machines
Every agent workflow has two roles: the orchestrator that decides, and the box where work runs. Split them and one driver runs many machines.
Blog
Thoughts on ephemeral compute, AI agents, and fire.
A Safer Box for the Sketchy Stuff
Every gibil server is already an isolated VM — your own kernel, your own IP, burned when you're done. We just sealed the channel into it.
What $300 of Vultr Credit Actually Buys You
Vultr gives new accounts $300 in free credit, valid for 30 days, via Gibil's referral. Here's how to actually use it — burn math, region picks, three workflows.
Gibil v0.4.0: Your Own Machine, Wherever You Are
Vultr support lands alongside Hetzner. Tokyo, Seoul, Singapore, Sydney, Mumbai — five new APAC regions, one CLI. New Vultr accounts get $300 in free credits.
Sandcastle, on Gibil: AFK Agents on Real Disposable Servers
A new Sandcastle sandbox provider that runs your Claude Code agents on fresh Hetzner servers instead of local Docker. Same orchestration, different box.
Gibil v0.3.2: Your Servers, Your Timeframe
Human-friendly TTL durations, a 1-year cap, and no more mental math. Here's what changed and why.
Move Your AI Agent Off Your Laptop
Even with one agent, your laptop shouldn't be doing the work. One command gets you a remote server instead of an afternoon of setup.
Gibil v0.2.0: Don't Switch Branches. Spin Up a Server.
New branch command, remote coding agents, port forwarding, and a security hardening pass. Here's what's new.
Run Your Test Suite on a Disposable Server
We ran Zod (3,574 tests) and Fastify (2,206 tests) on ephemeral ARM servers. Here's what we learned.
Designing CLIs for AI Agents: The --json Pattern
AI agents are the new power users. Here's how to build CLIs they can actually use.
Stop Polluting Your Local Machine
Your laptop is for thinking. Builds, tests, Docker, and experiments belong on a disposable server.
AI Agents + Gibil MCP: Development on Real Infrastructure
How AI agents use MCP tools to read, write, and run code on a remote server — without touching your machine.
VMs vs Sandboxes vs Cloud Sessions: Picking the Right Compute for AI Agents
Three architectural shapes for the box your agent runs in. Each makes a different tradeoff.
Fire That Forges, Fire That Fades: The Story Behind Gibil
Why we named a dev tool after a 5,000-year-old Sumerian fire god — and what a salamander has to do with it.
Why AI Agents Need Servers, Not Sandboxes
Sandboxes cap at 4 vCPU with no root and limited sessions. Agents doing real work need full machines.