Porta — DevOps skill for Claude Code
Sandboxed runtime for AI agents.
How to install Porta
This entry records only its repository, not the path inside it, so there is no
exact command to give. Open almide/porta and copy the folder into
~/.claude/skills/, or the file into ~/.claude/agents/.
What Porta does
Sandboxed runtime for AI agents. WASM isolation + OS-level sandbox. Run Claude Code, Python, or any command with filesystem/network restrictions. No Docker required.
Alternatives in DevOps
- Kubeshark — Cluster-wide network observability for Kubernetes 11.8k ★
- Mirrord — Run any process, on your machine or in an AI agent's environment, as if it were a pod in your Kubernetes clust 5.3k ★
- Azure Storage Blob Python — Azure Blob Storage for Python 1.8k ★
README
Porta
Run an agent with the permissions you actually granted it.
OS-enforced limits for the CLI agents you already use, and a WASM runtime for the ones you build.
Almide + Wasmtime · No Docker required
English · 日本語
Who this is for
- You run a CLI agent and want it unable to write outside one directory or reach hosts you did not allow — enforced by the OS, not by a prompt.
- You hand agents to a team and need a record of where they tried to connect.
- You build agents and need a run that can be resumed after a crash without repeating work, and a completion check the agent cannot talk its way past.
If none of these are your problem yet, Porta will not be interesting. It is a runtime and a set of restrictions, not an agent.
Install
bash scripts/install-almide.sh
.tools/almide/almide build src/mod.almd -o target/porta
cp target/porta ~/.local/bin/
Needs Almide 0.63.0, a Rust toolchain, Python 3 and curl. Full verification steps and the compiler pin are in [Build from source](#build-from-source).
Quick Start
1. Restrict an agent you already use
porta run claude --allow-net 'api.anthropic.com:443' -v ./project -e "HOME=$HOME" \
-- --print "Fix the bug in main.rs"
`claude` runs normally, but it can write only inside `./project` and connect only to the host you listed. No Docker daemon, no container image, no change to the agent itself.
2. See a restriction actually stop something
A working restriction is invisible, so watch one fail:
porta run curl -- https://example.com # network open by default
porta run curl --allow-net '*:443' -- https://example.com # HTTPS allowed →
Related Skills
Sandbox Shell
macOS Seatbelt sandbox CLI for developers. Protect credentials (SSH, AWS, GPG) from malicious npm packages, su
Sandy
Sandboxed TypeScript runtime for AI coding agents to query AWS — full SDK access with in-sandbox aggregation,
Claude Safe
Sandboxed Claude Code launcher — runs Claude in Docker with tmux session management and git worktree isolation
Cleat
Give the agent a cage, not your keys. One-command Docker sandbox for AI coding agents: full autonomous permiss
Ohmyharness
Run any coding agent in a Docker sandbox with your rules, skills and MCP servers already inside. Git worktree
Codejam Code Based Agents
Learn how to write code-based agents using Python, JavaScript and well-known industry standard frameworks. Lea
Related Agents
Tsk - AI Agent Task Manager And Sandbox
by dtormoen - A Rust CLI tool that lets you delegate development tasks to AI agents running in sandboxed Docke
Network Security Engineer
Network Security Engineer (Tier 3): hardens the network perimeter and runtime — TLS 1.3/HSTS/ACME, nginx/Caddy
Oneshot CI Fixer
Fix sandbox agent. Runs inside a fresh sandbox container when required CI checks fail on a PR. Reads the failu