by TM9657
Flow-Like: Strongly Typed Enterprise Scale Workflows. Built for scalability, speed, seamless AI integration and rich customization.
# Add to your Claude Code skills
git clone https://github.com/TM9657/flow-likeLast scanned: 5/4/2026
{
"issues": [],
"status": "PASSED",
"scannedAt": "2026-05-04T06:42:47.022Z",
"semgrepRan": false,
"npmAuditRan": false,
"pipAuditRan": true
}flow-like is an open-source ai agents skill for AI coding assistants such as Claude Code, Codex CLI, and ChatGPT, built by TM9657. Flow-Like: Strongly Typed Enterprise Scale Workflows. Built for scalability, speed, seamless AI integration and rich customization. It has 865 GitHub stars.
Yes. flow-like passed SkillsLLM's automated security scan — a dependency vulnerability audit plus prompt-injection heuristics — with no high-severity issues. You can read the full report in the Security Report section on this page.
Clone the repository with "git clone https://github.com/TM9657/flow-like" and add it to your Claude Code skills directory (see the Installation section above).
flow-like is primarily written in Rust. It is open-source under TM9657 on GitHub, so you can review or fork the full source.
Yes. SkillsLLM lists many other AI Agents skills you can browse and compare side by side. Open the AI Agents category from the badge at the top of this page, or use the Related Skills and comparison links further down to weigh flow-like against similar tools.
No comments yet. Be the first to share your thoughts!
Flow-Like is a visual workflow automation platform that runs entirely on your hardware. Build workflows with drag-and-drop blocks, run them on your laptop, phone, or server, and get a clear record of where data came from, what changed, and what came out — no cloud dependency, no black boxes, no guesswork.
Most workflow tools force you into their cloud. Your data leaves your machine, passes through third-party servers, and you're locked into their infrastructure. Offline? Stuck. Want to self-host? Pay enterprise prices.
Flow-Like runs wherever you choose — your laptop, your phone, your private server, or cloud infrastructure you control. Your data stays where you put it. No forced cloud dependency, no vendor lock-in, no "upgrade to enterprise for basic autonomy."
Local-first by default. Cloud-ready when you need it.
Automate on your couch without Wi-Fi. Deploy to your own AWS/GCP/Azure. Run in air-gapped factory networks. Process sensitive data in hospital environments. Your workflows, your infrastructure, your rules.
The reason this works is raw performance. Flow-Like's engine is built in Rust — compiled to native code, no garbage collector, no runtime overhead. The same workflow that takes 500ms in a Node.js engine takes 0.6ms in Flow-Like.
| Metric | Flow-Like | Typical workflow engines |
|---|---|---|
| Execution speed | ~244,000 workflows/sec | ~200 workflows/sec |
| Latency per workflow | ~0.6ms | ~50-500ms |
| Engine | Rust (native compiled) | Python / Node.js (interpreted) |
That 1000x performance gap means real workflows can run on resource-constrained devices — phones, edge hardware, Raspberry Pis — not just beefy cloud servers. And on powerful machines, it means processing millions of executions without breaking a sweat.
Most workflow tools show a green checkmark and move on. You're left guessing where data came from and why the result looks the way it does.
Flow-Like workflows are fully typed — they track what data flows where and why. Every input, transformation, and output is recorded with complete lineage and audit trails.
| Feature | Flow-Like | n8n | Zapier / Make | Temporal |
|---|---|---|---|---|
| Runs on your device | ✅ Desktop, phone, edge, server | ⚠️ Needs a server | ❌ Cloud only | ⚠️ Needs infrastructure |
| Works 100% offline | ✅ Full capability | ⚠️ Partial | ❌ Requires internet | ✅ Self-hosted |
| Type safety | ✅ Fully typed | ❌ Runtime only | ❌ None | ⚠️ Language-level |
| Data lineage / audit trail | ✅ Complete | ❌ Limited | ❌ None | ⚠️ Via logging |
| Performance | ✅ ~244K/sec (Rust) | ⚠️ ~200/sec (Node) | ⚠️ Cloud-limited | ⚠️ Go-based |
| Visual builder | ✅ Full IDE | ✅ Good | ✅ Simple | ❌ Code only |
| UI builder | ✅ Built-in | ❌ None | ❌ None | ❌ None |
| LLM orchestration | ✅ Built-in + guardrails | ⚠️ Via nodes | ⚠️ Via integrations | ❌ Manual |
| White-label / embed | ✅ Full customization | ❌ Branded | ❌ Branded | ❌ No UI |
| Business process views | ✅ Process / Data / Execution | ❌ Single view | ❌ Single view | ❌ Code only |
| License | Source Available (BSL) | Sustainable Use | Proprietary | MIT |
| 💻 Desktop App | ☁️ Web App | 📱 Mobile App | ⚙️ From Source |
|---|---|---|---|
| Download Now | Try Online | Coming Soon | Build Yourself |
| macOS · Windows · Linux | Available now | iOS · Android | Latest features |
A no-code IDE for building workflows. Smart wiring with type-aware pins, inline execution feedback, live validation, and snapshot-based debugging.
APIs & webhooks, databases, file processing (Excel, CSV, PDF), AI models & computer vision, messaging (Slack, Discord, email), IoT, logic & control flow, security & auth — and growing.
Download and run LLMs, vision models, and embeddings locally or in the cloud. Every AI decision is logged with full context — inputs, outputs, model version, and reasoning trace.
Package workflows as shareable applications with built-in storage. Run them offline or in the cloud. Browse the template store or share your own.
# Prerequisites: mise, Rust, Bun, Tauri prerequisites, Protobuf compiler
# Full guide: https://docs.flow-like.com/contributing/getting-started/
git clone https://github.com/TM9657/flow-like.git
cd flow-like
mise trust && mise install # install toolchains (Rust, Bun, Node, Python, uv)
bun install # install Node packages
mise run build:desktop # production build
All dev / build / deploy tasks are defined in the root mise.toml.
Run mise tasks to see every available task, or mise run <task> to execute one:
mise run dev:desktop:mac:arm # dev mode – macOS Apple Silicon
mise run dev:web # dev mode – Next.js web app
mise run build:desktop # production desktop build
mise run fix # auto-fix lint (clippy + fmt + biome)
mise run check # run all checks without fixing
💡 Platform-specific hints for macOS, Windows, and Linux are in the docs.
Embed the visual editor in your application, or run the engine headlessly behind the scenes.