https://www.fluid.sh/
$ fluid.sh
GitHub Discord
Claude Code for infrastructure. Debug, act, and audit everything
Fluid does on your infrastructure.
What does that mean?
Fluid is a terminal agent that do work on production infrastructure
like VMs/K8s cluster/etc. by making sandbox clones of the
infrastructure for AI agents to work on, allowing the agents to run
commands, test connections, edit files, and then generate
Infra-as-code like an Ansible Playbook to be applied on production.
Why not just use an LLM to generate Infrastructure-as-Code (IaC)?
LLMs are great at generating Terraform, OpenTofu, Ansible, etc. but
bad at guessing how production systems work. By giving access to a
clone of the infrastructure, agents can explore, run commands, test
things before writing the IaC, giving them better context and a place
to test ideas and changes before deploying. I got the idea after
seeing how much Claude Code has helped me work on code, I thought "I
wish there was something like that for infrastructure", and here we
are.
Why not just provide tools, skills, MCP server to Claude Code?
Safety. I didn't want CC to SSH into a prod machine from where it is
running locally (real problem!). I wanted to lock down the tools it
can run to be only on sandboxes while also giving it autonomy to
create sandboxes and not have access to anything else. Fluid gives
access to a live output of commands run (it's pretty cool) and does
this by ephemeral SSH Certificates. Fluid gives tools for creating
IaC and requires human approval for creating sandboxes on hosts with
low memory/CPU and for accessing the internet or installing packages.
Installation
Note: this is a terminal agent (like Claude Code) meant to be
installed on your local laptop/workstation
$ curl -fsSL https://fluid.sh/install.sh | bash
$ fluid
Built for where you already work
[~]
Sandbox Isolation
Clone VMs instantly. Test changes in isolation before touching
production.
ls
Context-Aware
Fluid explores your host first - OS, packages, CLI tools - then
adapts.
>>>
Full Audit Trail
Every command logged. Every change tracked. Review before production.
.yaml
Ansible Playbooks
Auto-generates playbooks from sandbox work. Reproducible
infrastructure.
Usage
$ |
Creating sandbox...
v create_sandbox
ID: SBX-demo1234
IP: 192.168.122.50
State: RUNNING
Running: cat /etc/os-release
v run_command
exit: 0
NAME="Ubuntu"
VERSION="22.04.3 LTS (Jammy Jellyfish)"
Running: apt update
v run_command
exit: 0
Hit:1 http://archive.ubuntu.com/ubuntu jammy InRelease
Get:2 http://archive.ubuntu.com/ubuntu jammy-updates InRelease [119
kB]
Reading package lists... Done
... (2 more lines)
Running: apt install -y apache2
v run_command
exit: 0
Reading package lists... Done
Building dependency tree... Done
The following NEW packages will be installed:
... (7 more lines)
Running: echo '...' > /var/www/html/index.html
v run_command
exit: 0
Running: systemctl status apache2
v run_command
exit: 0
* apache2.service - The Apache HTTP Server
Loaded: loaded (/lib/systemd/system/apache2.service; enabled)
Active: active (running) since Sat 2026-01-25 12:00:00 UTC
Running: curl localhost
v run_command
exit: 0
Fluid Demo
Hello fr...
Creating playbook...
v create_playbook
Playbook: httpd-setup
Adding playbook tasks...
v add_playbook_task x4
Playbook: httpd-setup
Tasks: TASK-001, TASK-002, TASK-003, TASK-004
Done! Here's what I accomplished:
Sandbox Created
- ID: SBX-demo1234
- IP: 192.168.122.50
Apache HTTP Server
- Installed and running
- Custom page at /var/www/html/index.html
- Verified working with curl
Ansible Playbook: httpd-setup
The playbook includes 4 tasks:
1. Update apt cache
2. Install Apache
3. Create custom index.html
4. Start and enable Apache service
You can run this playbook on any Ubuntu server to reproduce this
setup.
(c) 2026 fluid.sh
GitHub Discord