Agentic Coding Flywheel Setup (ACFS)
One command to transform a fresh Ubuntu VPS into a full multi-agent AI development environment in about 30 minutes.
- Stars
- ★ 1.7k
- Last updated
- today
- License
- NOASSERTION
- Primary language
- Shell
- FA score
- 49/100 · Major gaps
At a glance
- How it runs
- Works with
- Platform-specificCodex · Claude Code
- Cost
- Free software; you pay for model usage
- Setup effort
- Medium · a few setup steps
- You'll need
- Typical use
- Individual developers wanting to quickly spin up a VPS for AI-assisted coding without manual setup.
- Not a fit if
- Users who need non-Ubuntu systems or want to avoid automatic OS upgrades and reboots
- Teams that want to avoid lock-in to specific agent CLIs and the Dicklesworthstone stack
- Users without a VPS of at least 4GB RAM and 10GB disk
- Source review
- 49/100 · Major gaps
What does this agent do, and when should you use it?
ACFS is a Bash bootstrapping system that converts a fresh Ubuntu VPS (22.04+, auto-upgrades to 25.10) into a complete development environment with three AI coding agents (Claude Code, Codex CLI, Antigravity CLI). It installs 30+ tools via a single curl|bash command, including shell environment (zsh, oh-my-zsh, powerlevel10k), language runtimes (bun, uv, Rust, Go), cloud CLIs (Vault, Wrangler, Supabase, Vercel), and the Dicklesworthstone stack (ntm, ubs, etc.). The installer is idempotent, resumable, and includes preflight and checksum-based verification. The project also includes an interactive wizard website (agent-flywheel.com), a terminal TUI onboarding, a learning hub, and the ACFS CLI for environment management.
ACFS executes installation scripts in phases on the target VPS: user normalization, APT packages, shell setup, CLI tools, language runtimes, AI agents, cloud tools, Dicklesworthstone stack, configuration deployment, and verification. It generates and deploys configuration files (~/.acfs/zshrc, tmux.conf), creates aliases like cc, cod, agy, and generates acfs doctor health checks, acfs info system overview, and acfs update update commands. It uses acfs.manifest.yaml as a single source of truth for all tool definitions, generating category installers and doctor_checks.sh from the manifest. The installer supports 'vibe' and 'safe' modes for passwordless sudo and agent permission skipping (vibe) or standard confirmations (safe).
- Individual developers wanting to quickly spin up a VPS for AI-assisted coding without manual setup.
- Beginners needing step-by-step guidance through the interactive wizard (agent-flywheel.com).
- Teams needing a reproducible and consistent development environment for agentic workflows.
- Users wanting to leverage the pre-installed Dicklesworthstone stack (ntm, ubs, etc.) for multi-agent coordination.
- Users wanting to customize tool installations via a centralized YAML manifest.
How do you install or deploy this agent?
Requires a fresh Ubuntu VPS (22.04+) with SSH access. On the VPS, run: curl -fsSL "https://raw.githubusercontent.com/Dicklesworthstone/agentic_coding_flywheel_setup/main/install.sh?$(date +%s)" | bash -s -- --yes --mode vibe. The installer will automatically upgrade Ubuntu to 25.10 (can be skipped) and install all tools. Before running the installer, you can run a preflight check (scripts/preflight.sh) to validate the system. After installation, configure agent credentials with acfs services-setup (Claude API key, Codex login, etc.).
How do you use this agent?
After installation, use the three AI agents directly via aliases: cc (Claude Code), cod (Codex), and agy (Antigravity). Use acfs newproj to create a new project with AGENTS.md, acfs doctor for health checks, and acfs update to update all tools. Launch the interactive tutorial with onboard, and access the learning hub at agent-flywheel.com/learn. Example: on your VPS, in a project directory, run cc to start a Claude Code session.
What are this agent's strengths and limitations?
- One-shot installation of 30+ tools and three AI agents, saving significant manual configuration time
- Manifest-driven generation ensures consistency between install logic and documentation
- Security features include preflight checks, checksum verification, HTTPS enforcement, and content hash checking
- Idempotent and resumable installation that automatically continues from the last completed phase
- Ubuntu-only (22.04+) and auto-upgrades to 25.10, which can involve reboots and significant time
- Tightly coupled to the Dicklesworthstone stack and specific agent CLIs (Claude, Codex, Antigravity), creating vendor lock-in
- 'Vibe' mode enables dangerous flags (e.g., --dangerously-skip-permissions), posing security risks
- Requires substantial VPS resources (at least 4GB RAM, 10GB disk)
How does this agent compare with similar options?
Key facts side by side with the most closely related agents.
| Agent | Source review | Form / cost | Stars | Updated | Language | Full support on |
|---|---|---|---|---|---|---|
| Agentic Coding Flywheel Setup (ACFS) This agent | 49 · Major gaps | CLIFree + model costs | ★ 1.7k | today | Shell | Codex · Claude Code |
| Claude Council | 74 · Some gaps | Agent plugin / skillFree + model costs | ★ 890 | 1d ago | Shell | Claude Code · OpenAI API · Claude API |
| CCB Multi-Agent Terminal Workspace | 64 · Some gaps | CLIFree + model costs | ★ 3.6k | 5d ago | Python | Codex · Claude Code |
| AI Switcher | 58 · Major gaps | CLIFree | ★ 133 | 4d ago | Rust | Codex · Claude Code |
How does FollowAgents rate this agent?
Why each dimension lost points
Evidence shows: installer offers safe mode and vibe mode; safe mode disables passwordless sudo and requires confirmation, but default recommended vibe mode enables passwordless sudo and dangerous flags, and installation via curl|bash skips user confirmation in non-interactive mode. Data flow transparency: README describes installation flow and state file, but lacks details on data collection or transmission. Sensitive data handling: involves SSH key migration, but key storage and permission details are not specified. Dependency security: implements checksums.yaml and SHA256 verification, but auto-update mechanism may introduce risk. External effects: installs many tools and modifies system configuration, but no rollback mechanism provided. Rollback: only re-install and state recovery, no true rollback. Source attribution: author and repository clearly identified, but publisher unverified. Deductions: default recommendation of vibe mode weakens least privilege and user confirmation; data flow and sensitive data handling details insufficient; rollback capability limited.
Evidence shows: README and code structure consistent; installer designed idempotent and resumable, but no actual test results provided. Dependency availability: depends on external installers and network, but provides preflight and cache mechanisms; actual availability not verified. Failure messages: preflight error table and logs provided, but detailed installation failure diagnostics not provided. Deductions: external dependencies, static review cannot verify availability; failure messages exist but not comprehensive.
Evidence shows: targets beginners, developers, and teams; provides wizard website and multiple installation modes, but capability boundaries unclear (e.g., which tools are optional). Trigger precision: installer controlled via command-line arguments, but not all options documented. Environment fit: supports Ubuntu 22.04+, but other systems not mentioned. Deductions: capability boundaries and trigger options documentation incomplete.
Evidence shows: README structure clear, includes installation instructions, architecture diagrams, usage examples, but lacks FAQ and known limitations. Naming stability: version number exists, but naming rules not specified. License: MIT+OpenAI/Anthropic Rider, but metadata NOASSERTION. Versioning: version number and changelog exist, but detailed change records not provided. Maintenance responsibility: author and repository identified, but maintenance commitment not explicit. Deductions: missing FAQ, known limitations, and detailed changelog.
Evidence shows: output is usable development environment, provides 30+ tools and three AI agents, high marginal value, cost-benefit reasonable (free open source). Deductions: static review cannot verify actual effectiveness.
Evidence shows: README claims partially supported by code (e.g., checksums.yaml), but no independent verification. Cross-source: relies solely on repository internal evidence. Fact/inference separation: README distinguishes descriptions and inferences, but not clearly. Deductions: lack of external verification and clear fact/inference separation.
- Default recommended vibe mode enables passwordless sudo and dangerous agent flags, posing security risks.
- Installer executes via curl|bash, skipping user confirmation in non-interactive mode.
- Auto-updating checksums.yaml may introduce unreviewed upstream changes.
- Lack of rollback mechanism makes recovery difficult after failed installation.
- License includes OpenAI/Anthropic Rider, potentially restricting certain use cases.
Recent releases
Latest entries from the project's GitHub Releases. Notes are quoted from upstream as written, not rewritten or tested by FollowAgents.
-
Release notes excerpt
Public copies of vendor-verified CLI archives for restricted cloud environments. The cloud-mirror.json manifest pins each artifact by SHA256 and retains its upstream provenance. Assets are content addressed and never replaced. This is not the latest ACFS software release.View on GitHub ↗ -
v0.10.0 v0.10.0 (source only)
Release notes excerpt
ACFS v0.10.0 preserves supported Ubuntu 22.04/24.04 LTS hosts on ordinary installs. An OS release upgrade now requires an explicit target, and --skip-ubuntu-upgrade takes precedence. Non-root sudo users can validate completed root-only checkpoints without changing their permissions. These changes are on both main and master. This is a source-only installer release, matching the existing ACFS publication venue. It has no uploaded binary assets. …View on GitHub ↗ -
v0.9.0 v0.9.0 — service repair, per-tool holds with rollback, Gemini 3.8 Antigravity
Release notes excerpt
ACFS v0.9.0 ACFS is installed with curl | bash, so this release is the script itself, not a set of binaries: there are intentionally no downloadable assets. The deliverable is that this tag installs. > 57 commits since v0.8.0 (2026-08-25 through 2026-09-04). Installer-only release, like every ACFS release: the deliverable is that this tag installs. Pinned install: curl -fsSL "https://raw.githubusercontent.com/Dicklesworthstone/agentic_coding_flywheel_setup/v0.9.0/install.sh" | bash -s -- --yes --mode vibe --ref v0.9.0. Update & service reliability …View on GitHub ↗