Isolated machines for AI coding agents - with active defense.

COI gives each AI agent its own machine - a full system container with root access, systemd, Docker, and the ability to install anything. Agents work like they would on a real server: run services, manage packages, use cron - without touching your actual system. Files stay correctly owned, no permission hacks needed.

Your credentials stay on the host. SSH keys, environment variables, and Git tokens are never exposed to AI tools unless you explicitly mount them. If something goes wrong, COI catches it - reverse shells, credential scanning, data exfiltration - and pauses or kills the container automatically. No manual intervention needed.

Built by developers, for developers who run AI agents and want to know what those agents are doing. Not a product, not a startup - a tool that does the job.

Who this is for

• You run AI coding agents and want them to have full machine access - root, Docker, package managers, services - without risking your host
• You want to know when an agent does something suspicious, not find out after the fact
• You run multiple agents in parallel and need them isolated from each other
• You want persistent dev environments that survive restarts and reboots, not throwaway containers that lose your setup every time
• You care about your credentials not ending up inside an agent-controlled environment

Watch the BetterStack video about Code on Incus

Table of Contents

• Who this is for
• Supported AI Coding Tools
• Supported Tools (detailed)
• Features
• Quick Start
• Why Incus Instead of Docker or Docker Sandboxes?
• Installation
• macOS Support
• Usage
• Session Resume
• Persistent Mode
• Configuration
• Profiles
• Resource and Time Limits
• Container Lifecycle & Session Persistence
• Network Isolation
• Security Monitoring
• Security Best Practices
• Snapshot Management
• System Health Check
• Troubleshooting
• FAQ

Supported AI Coding Tools

Currently supported:

• Claude Code (default) - Anthropic's official CLI tool
• opencode - Open-source AI coding agent (https://opencode.ai)

Coming soon:

• Aider - AI pair programming in your terminal
• Cursor - AI-first code editor
• And more...

Tool selection:

coi shell                    # Uses default tool (Claude Code)
coi shell --tool opencode    # Use opencode instead

Permission mode - Control whether AI tools run autonomously or ask before each action:

# ~/.coi/config.toml or .coi/config.toml
[tool]
name = "claude"              # Default AI tool
permission_mode = "bypass"   # "bypass" (default) or "interactive"

See the Supported Tools wiki page for detailed configuration, API key setup, and adding new tools.

Features

Core Capabilities

• Multi-slot support - Run parallel AI coding sessions for the same workspace with full isolation
• Session resume - Resume conversations with full history and credentials restored (workspace-scoped)
• Persistent containers - Keep containers alive between sessions (installed tools preserved)
• Workspace isolation - Each session mounts your project directory
• Slot isolation - Each parallel slot has its own home directory (files don't leak between slots)
• Workspace files persist even in ephemeral mode - Only the container is deleted, your work is always saved
• Container snapshots - Create checkpoints, rollback changes, and branch experiments with full state preservation

Host Integration

• SSH agent forwarding - Use git-over-SSH inside containers without copying private keys ([ssh] forward_agent = true)
• Environment variable forwarding - Selectively forward host env vars by name (forward_env in config)
• Host timezone inheritance - Containers automatically inherit the host's timezone (configurable via [timezone] config)
• Sandbox context file - Auto-injected ~/SANDBOX_CONTEXT.md tells AI tools about their environment (network mode, workspace path, persistence, etc.). Automatically loaded into each tool's native context system: Claude Code via ~/.claude/CLAUDE.md, OpenCode via the instructions field in opencode.json (opt out with auto_context = false)

Security & Isolation

• Credential protection - SSH keys, .env files, Git credentials, and environment variables are never exposed unless explicitly mounted
• Privileged container guard - Refuses to start when security.privileged=true is detected, which defeats all container isolation
• Security posture verification - coi health checks seccomp, AppArmor, and privilege settings to confirm full isolation
• Kernel version enforcement - Warns on host kernels below 5.15 that may lack security features for safe isolation
• Real-time threat detection - Kernel-level nftables monitoring detects reverse shells, C2 connections, data exfiltration, DNS tunneling, and credential scanning
• Automated response - Auto-pause on HIGH threats, auto-kill on CRITICAL — no manual intervention needed
• Network isolation - nftables-based restricted/allowlist/open modes block private network access and prevent exfiltration
• Protected paths - .git/hooks, .git/config, .husky, .vscode mounted read-only to prevent supply-chain attacks
• Host-side immutable protection - Protected paths are locked with chattr +i during sessions, preventing unshare -m + umount bypass of read-only mounts (opt out: [security] host_immutable = false)
• Git identity guard - Containers enforce user.useConfigOnly=true, preventing AI tools from committing as the default "code" user
• Guest API disabled - Incus guest API (/dev/incus) disabled by default, preventing host path and topology leaks
• System containers - Full OS isolation with unprivileged containers, better than Docker privileged mode
• Automatic UID mapping - No permission hell, files owned correctly
• Audit logging - All security events logged to JSONL for forensics and compliance

Safe Dangerous Operations

• AI coding tools often need broad filesystem access or bypass permission checks
• These operations are safe inside containers because the "root" is the container root, not your host system
• Containers are ephemeral - any changes are contained and don't affect your host
• This gives AI tools full capabilities while keeping your system protected

Quick Start

# Install
curl -fsSL https://raw.githubusercontent.com/mensfeld/code-on-incus/master/install.sh | bash

# Build image (first time only, ~5-10 minutes)
coi build

# Start coding with your preferred AI tool (defaults to Claude Code)
cd your-project
coi shell

# Or use opencode instead
coi shell --tool opencode

# That's it! Your AI coding assistant is now running in an isolated container with:
# - Your project mounted at /workspace
# - Correct file permissions (no more chown!)
# - Full Docker access inside the container
# - GitHub CLI available for PR/issue management
# - All workspace changes persisted automatically
# - No access to your host SSH keys, env vars, or credentials

Why Incus Instead of Docker or Docker Sandboxes?

Incus is a modern Linux container and virtual machine manager, forked from LXD. Unlike Docker (which uses application containers), Incus provides system containers that behave like lightweight VMs with full init systems.

Security Comparison

Why Incus Instead of Docker Sandboxes?

• Linux-first, not Linux-last. Docker Sandboxes' microVM isolation is only available on macOS and Windows. Linux gets a legacy container-based fallback. COI is built for Linux from the ground up because Incus is Linux-native.

• No Docker Desktop required. Docker Sandboxes is a Docker Desktop feature. Docker Desktop is not open source and has commercial licensing requirements for larger organizations. COI depends only on Incus - fully open source, no vendor lock-in, no additional runtime.

• System containers, not containers-in-VMs. Incus system containers run a full OS with systemd and native Docker support inside - one clean isolation layer. Docker Sandboxes nests application containers inside microVMs, adding architectural complexity.

• No permission hell. Incus automatic UID/GID shifting means files created by agents have correct ownership on the host. No mapping hacks needed. (Note: files created via sudo in the workspace will be root-owned — the sandbox context file instructs AI tools to fix ownership after sudo operations.)

• Credential isolation by default. Host environment variables, SSH keys, and Git credentials are never exposed to AI tools unless explicitly mounted.

• Simple and transparent. No separate daemon, no opaque VM nesting. COI talks directly to Incus - easy to inspect, debug, and extend.

Installation

Automated Installation (Recommended)

# One-shot install
curl -fsSL https://raw.githubusercontent.com/mensfeld/code-on-incus/master/install.sh | bash

# This will:
# - Download and install coi to /usr/local/bin
# - Check for Incus installation
# - Verify you're in incus-admin group
# - Show next steps

Manual installation: Download the binary from GitHub Releases, make it executable, and move to /usr/local/bin/. Requires Linux with Incus installed and user in the incus-admin group. You must log out and back in (or run newgrp incus-admin) after adding your user to the group — COI runs incus directly and requires the group to be active in your session. See the Incus installation guide for setting up Incus.

Build Images

# Build the default coi-default image (5-10 minutes)
coi build

# Build without compression (faster iteration)
coi build --compression none

# Build a custom image via a profile
coi profile create my-image --image my-image
# Edit .coi/profiles/my-image/config.toml to add a [container.build] section
coi build --profile my-image

What's included in the coi-default image:

• Ubuntu 22.04 base with Docker (full Docker-in-container support)
• mise (polyglot runtime manager) — Python 3, pnpm, TypeScript, tsx pre-installed; add more with mise use go@latest, mise use ruby@3, etc.
• Node.js 20 LTS (system, for Claude CLI) + npm
• Claude Code CLI (default AI tool) + GitHub CLI (gh)
• tmux, git, curl, build-essential, and common build tools
• Modern CLI utilities: fd-find, bat, tree
• Debugging tools: strace, lsof
• Database clients: sqlite3, postgresql-client, redis-tools
• imagemagick for image processing

Custom images: Build your own specialized images using profile-based build scripts that run on top of the base coi-default image. See the Image Management wiki page for complete profile-based build workflows.

macOS Support

COI works on macOS using Colima or Lima VMs. See the macOS Setup Guide for complete instructions.

Usage

Basic Commands

# Interactive session (defaults to Claude Code)
coi shell

# Use a different AI tool
coi shell --tool opencode

# Persistent mode - keep container between sessions
coi shell --persistent

# Use specific slot for parallel sessions
coi shell --slot 2

# Resume previous session
coi shell --resume

# Run a command in an ephemeral container
coi run "npm test"

# Attach to existing session
coi attach

# Real-time security monitoring dashboard
coi monitor

# View session logs (setup messages, network notices, errors)
coi logs                        # Auto-detect container from current workspace
coi logs coi-abc123-1 -f        # Tail logs live

# List active containers and saved sessions
coi list --all

# Gracefully shutdown / force kill containers
coi shutdown coi-abc12345-1
coi kill --all

# Cleanup stopped containers and orphaned resources
coi clean
coi clean --pools             # Detect containers in unused storage pools

# Update coi to the latest release
coi update

Container Aliases

Assign human-friendly names to containers for easy management from any directory:

# .coi/config.toml (in your project)
[container]
alias = "myproject"

coi shell myproject              # Launch session using alias (from any directory)
coi attach myproject             # Attach to running aliased container

See the Container Lifecycle and Sessions guide for full alias documentation.

Global Flags

--workspace PATH        # Workspace directory to mount (default: current directory)
--slot NUMBER           # Slot number for parallel sessions (0 = auto-allocate)
--persistent            # Keep container between sessions
--resume [SESSION_ID]   # Resume from session (omit ID to auto-detect latest for workspace)
--continue [SESSION_ID] # Alias for --resume
--profile NAME          # Use named profile
--image NAME            # Use custom image (default: coi-default)

Most container customization (network mode, mounts, environment variables, SSH agent, monitoring, timezone, resource limits, etc.) is configured via config files or profiles. See the Configuration wiki page for the full reference.

Advanced Usage

See the wiki for detailed documentation:

• Container Operations - Container management and low-level operations
• File Transfer - Push/pull files between host and containers
• Tmux Automation - Automate AI sessions with tmux commands
• Image Management - Create and manage custom images
• Snapshot Management - Create checkpoints and rollback changes

Session Resume

Resume a previous AI coding session with full history and credentials restored:

coi shell --resume              # Auto-detect latest session for this workspace
coi shell --resume=<session-id> # Resume specific session
coi list --all                  # List available sessions

What's restored: Full conversation history, tool credentials, user settings, and project context. The profile used when the session was created is also automatically restored — no need to pass --profile again (explicitly passing --profile overrides the saved one). Sessions are workspace-scoped — --resume only finds sessions from the current workspace directory.

See the Container Lifecycle and Sessions guide for details on how session persistence works.

Persistent Mode

By default, containers are ephemeral (deleted on exit). Your workspace files always persist regardless of mode.

Enable persistent mode to also keep the container and its installed packages:

# Or via config (~/.coi/config.toml)
[container]
persistent = true

What persists:

• Ephemeral mode: Workspace files + session data (container deleted)
• Persistent mode: Workspace files + session data + container state + installed packages, system setup

See the Container Lifecycle and Sessions guide for details.

Configuration

Config file: ~/.coi/config.toml

[container]
image = "coi-default"
persistent = true
# storage_pool = ""            # Empty = Incus default pool
# alias = "myproject"          # Human-friendly name for this workspace's containers

[tool]
name = "claude"
permission_mode = "bypass"
# auto_context = true          # Auto-inject sandbox context into tool's native system

Configuration hierarchy (highest precedence last):

1. Built-in defaults
2. User config (~/.coi/config.toml)
3. Project config (./.coi/config.toml)
4. COI_CONFIG environment variable
5. Environment variables (CLAUDE_ON_INCUS_*, COI_*)
6. Operational CLI flags (--workspace, --slot, --persistent, --resume, --profile, --image)

Place a .coi/config.toml in any repository root to auto-configure COI for that project — useful for teams to share container image, environment, and resource limits.

See the Configuration wiki page for the full config reference, per-repo setup, profiles, and environment variables.

Profiles

Profiles are reusable container configurations bundling image, tool, limits, mounts, build scripts, context files, and environment into named templates. Each profile is a self-contained directory under profiles/:

.coi/profiles/
├── rust-dev/
│   ├── config.toml      # profile config
│   ├── build.sh         # profile-specific build script
│   └── CONTEXT.md       # AI agent context (appended to sandbox context)
└── python-ml/
    ├── config.toml
    └── setup.sh

Example profile config (.coi/profiles/rust-dev/config.toml):

context = "CONTEXT.md"
forward_env = ["CARGO_HOME"]

[container]
image = "coi-rust"
persistent = true

[environment]
RUST_BACKTRACE = "1"

[tool]
name = "claude"
permission_mode = "bypass"

[limits.cpu]
count = "4"

coi shell --profile rust-dev                          # Use a profile
coi profile create rust-dev --image coi-rust           # Create a new profile
coi profile list                                       # List all profiles

Profiles support inheritance (inherits = "parent-name"), context files for AI-agent instructions, and custom build scripts. See the Profiles wiki page for complete documentation.

Profile JSON Schema

COI ships a JSON Schema 2020-12 document that describes every field accepted by a profile config.toml. External tools — such as a web UI or editor plugin — can consume it to validate profile data without duplicating COI's validation logic. The self-contained schema (with all definitions bundled inline) is produced by coi schema profile; the source files live under schema/ but are incomplete without the bundling step.

# Print the schema
coi schema profile

# Save it for use in another tool
coi schema profile > profile.schema.json

The schema covers all field types, enum values (network.mode, tool.permission_mode, tool.claude.effort_level, timezone.mode), required fields on mount entries, and rejects unknown keys. Any JSON Schema 2020-12 validator can use it — for example, the Ruby json_schemer gem or the Python jsonschema package.

Resource and Time Limits

See the Resource and Time Limits guide for complete documentation on controlling container resource consumption and runtime.

Quick example:

# ~/.coi/config.toml
[limits.cpu]
count = "2"

[limits.memory]
limit = "2GiB"

[limits.runtime]
max_duration = "2h"

What you can limit:

• CPU cores and usage percentage
• Memory and swap
• Disk I/O rates
• Maximum runtime and process count
• Auto-stop on time limits

Container Lifecycle & Session Persistence

See the Container Lifecycle and Sessions guide for detailed explanation of how containers and sessions work.

Key concepts:

• Workspace files: Always saved (regardless of mode)
• Session data: Always saved to ~/.coi/sessions-<tool>/
• Ephemeral mode (default): Container deleted after exit, session preserved
• Persistent mode (--persistent): Container kept with all installed packages
• Resume (--resume): Restore AI conversation in fresh/existing container

Quick reference:

coi shell --persistent        # Keep container between sessions
coi shell --resume            # Resume previous conversation
coi attach                    # Reconnect to running container
coi persist                   # Convert ephemeral session to persistent
coi unfreeze <name>           # Unfreeze paused/frozen container
coi unfreeze                  # Unfreeze all frozen COI containers
close                         # Properly stop container (inside, safe alias for poweroff)
coi shutdown <name>           # Graceful stop (outside)

Network Isolation

See the Network Isolation guide for complete documentation on network security and nftables-based network filtering.

Network modes:

• Restricted (default) - Blocks private networks, allows internet
• Allowlist - Only specific domains/IPs allowed
• Open - No restrictions (trusted projects only)

# ~/.coi/config.toml
[network]
mode = "restricted"   # Default — blocks private networks, allows internet
# mode = "allowlist"  # Only specific domains/IPs allowed
# mode = "open"       # No restrictions (trusted projects only)

Security Monitoring

COI includes built-in security monitoring to detect and respond to malicious behavior in real-time:

# Enable in config (~/.coi/config.toml)
[monitoring]
enabled = true

Protects against:

• Reverse shells - Detects common reverse shell patterns (auto-kill)
• Data exfiltration - Monitors large workspace reads/writes (auto-pause)
• Environment scanning - Flags processes searching for API keys and secrets
• Network threats (NFT) - Kernel-level detection of C2 connections, private network access, DNS tunneling, and allowlist violations

Automated response levels:

• INFO/WARNING: Logged (+ alert for WARNING)
• HIGH: Container paused (requires coi unfreeze to continue)
• CRITICAL: Container killed immediately

Audit logs are stored at ~/.coi/audit/<container-name>.jsonl in JSON Lines format.

See the Security Monitoring wiki page for monitoring commands, configuration options, NFT setup, and audit log management.

Audit streaming

coi audit exposes the audit stream as JSON Lines on stdout, ready to pipe into a SIEM, jq, or a flat file:

# Dump the host-side audit log for a container
coi audit coi-abc-1

# Live in-container collector: auditd if available, otherwise syslog/auth.log,
# plus periodic ss + ps snapshots
coi audit coi-abc-1 --follow

# Re-stream a saved JSONL file
coi audit --file ./session.jsonl | jq -c 'select(.type=="net")'

Each line is a JSON object of the form:

{"ts":"2026-05-05T12:34:56.789Z","sessionId":"coi-abc-1","container":"coi-abc-1",
 "type":"exec|net|file|audit|heartbeat","pid":42,"comm":"curl","args":"curl https://...",
 "peer":"1.2.3.4:443","path":"/etc/shadow","msg":"...","raw":"..."}

The --follow collector is a small POSIX-sh agent (agent.sh, no binary deploy). It picks the best available source — auditd first, falling back to tail -F of /var/log/syslog and /var/log/auth.log — and adds ss -tunp snapshots every 5 s plus ps-tree diffs every 2 s for new pids. Measured idle overhead: ~4.5 MB total RSS across the agent and its tail/awk helpers, ~0.0% CPU when the container is quiet.

Heartbeat / silent-failure detection. The agent emits a {"type":"heartbeat","seq":N,"sources":"..."} event every 10 s. The host-side watcher tracks the most recent heartbeat per session and surfaces a warning if none arrives for 35 s (3 missed heartbeats with a small grace margin), both on stderr:

[audit] WARNING agent silent on coi-abc-1 for 36s (last heartbeat 2026-05-05T12:34:46Z)

and inside the JSONL stream as a type=audit msg=agent.stale event so SIEMs and downstream tooling can pick it up. When heartbeats resume, an agent.alive event clears the warning. The watcher does not auto-kill the connection — operators decide. Tunables: COI_AUDIT_HEARTBEAT_INTERVAL on the agent side; the 35 s stale threshold and 5 s check cadence are host-side defaults exposed as audit.DefaultStaleAfter and audit.DefaultCheckInterval for embedders.

Privacy model: every event stays on the host running coi audit. Nothing is sent off-machine unless you pipe it elsewhere yourself.

Security Best Practices

See the Security Best Practices guide for detailed security recommendations.

COI automatically mounts security-sensitive paths as read-only to prevent supply-chain attacks:

• .git/hooks, .git/config, .husky, .vscode

Set [git] writable_hooks = true in config to opt out, or customize protected paths via config. See the wiki for details.

System Health Check

See the System Health Check guide for detailed information on diagnostics and what's checked.

Run diagnostics:

coi health                    # Basic health check
coi health --format json      # JSON output
coi health --verbose          # Additional checks

What it checks: System info, kernel version, Incus setup, permissions, security posture (seccomp/AppArmor), privileged container detection, network configuration, storage, monitoring prerequisites, and running containers.

Exit codes: 0 (healthy), 1 (degraded), 2 (unhealthy)

Troubleshooting

See the Troubleshooting guide for common issues and solutions.

Common issues:

• DNS issues during build - COI automatically fixes systemd-resolved conflicts
• Run coi health to diagnose setup problems
• Check the troubleshooting guide for detailed solutions

Frequently Asked Questions

See the FAQ for answers to common questions.

Topics covered:

• Orphaned nftables/iptables rules
• How COI compares to Docker Sandboxes and DevContainers
• Windows support (WSL2)
• Security model and prompt injection protection
• API key security and trust model
• What is Incus? (vs tmux)

Getting Help

• Slack: Join the COI community on Slack — ask questions, report issues, share feedback
• GitHub Issues: Open an issue for bug reports and feature requests
• Wiki: Browse the documentation wiki for guides and reference