Rahul KumarAI creator
AI coding agents can build an entire feature in minutes. But do you actually understand the system they just built? Instead of starting with more code, it starts with the codebase.
Systems-design-first coding harness
Generate architecture diagrams
straight from your codebase
Point it at your repo and it draws the architecture, then hands that diagram to the coding agent as a specification. Desktop app or a 6 MB Rust binary. Your models, fully offline.
From the founderSystems Design using AI Agents
How it works
The diagram is the specification
One request goes through a small orchestrator, comes back as an architecture drawing, and leaves as the spec the coding agent builds from.
Multi-project workspaces
Several repos at once, each with its own agent sidecar and history.
Reviewable diffs & approvals
Every edit lands as a diff. Commands are risk-tiered; catastrophic is refused.
Command palette
Tabs, projects, theme and your editor, all from ⌘K.
MCP & skills
Any stdio MCP server, plus SKILL.md packs from skills.sh.
Voice & vision
Push-to-talk with on-device transcription. Attach images for vision models.
Light, dark & your editor
One identity in both schemes. Opens VS Code, Cursor or Zed in a click.
Design mode
Describe a page, get self-contained HTML. Click one element to rewrite only it.
Dependency & MCP audit
osv-scanner, npm, cargo and pip audit over nine ecosystems, plus your MCP config.
Grafana in the window
Grafana's official MCP server for dashboards, Prometheus and Loki.
The walkthrough
The architecture diagram generator, end to end
Point Lucid Train at a repository and watch it draw the architecture, then hand that diagram to the coding agent as a specification. Nothing sped up, nothing edited around.
Arnab Bhattacharyawatch on YouTube
The desktop app
One app. Every surface
Chat with the coding agent, sketch architecture, audit dependencies, watch Grafana, wire up connectors.
↑ click the tabs · Code · Audit · Dashboards · Connectors · Diagram
What people are saying
Shruti Codes@Shruti_0810
Stop asking AI to write more code. Start asking it to understand the code you already have. You have thousands of files, hundreds of dependencies, and no idea what connects to what.
Arnab Bhattacharya@Arnab28122000the author
I was not sure what claude or codex wrote during my vibe coding sessions. So I developed an app that can be used to visualize the architecture of the not so sloppy code.
>_ The CLI
Prefer the terminal?
So do we
The same Terminal-Bench-grade harness, as a single 6 MB Rust binary with a fast, keyboard-first TUI. Live plans, thinking stream, diffs, token accounting. No IDE, no browser.
$curl -fsSL https://lucidtrain.com/install.sh | sh
/agent full tools
/plan propose-only
/research cited, no shell
Models · open models first
Your models. Your machine
Any OpenAI-compatible endpoint works. Run fully local and free, or bring your own key. Your prompts go straight to the provider you pick, never through us.
Local · free
Ollama, zero-setup
/download installs Ollama and pulls the best model for your RAM. 100% offline.
gpt-oss:20b
qwen3-coder:30b
deepseek-r1
qwen3-coder:30b
deepseek-r1
OpenRouter · BYO key
15+ models
Bring your key. Stored locally, sent only to the provider.
GPT · Gemini · Grok
DeepSeek V4 · Qwen
GLM · MiniMax
DeepSeek V4 · Qwen
GLM · MiniMax
Subscriptions · no key
Already paying?
login claude or login codex to use the plan you already have, no API key.
Claude Pro / Max
ChatGPT Plus / Pro
via the official CLIs
ChatGPT Plus / Pro
via the official CLIs
Pricing
The CLI is free. The desktop app is $3
One licence for the desktop app. Everything runs on your machine, against local models or your own API keys, so there is no usage to meter and nothing to top up.
Terminal
CLI
Free · forever
- The full agent, in one Rust binary
- Local models, or your own API keys
- Open source · no account needed
$curl -fsSL https://lucidtrain.com/install.sh | sh
or: pip install lucid-train
Desktop app
2 months freeLucid Train
$2.50/mo
billed $30 a year
- Everything in the CLI, in a native app
- Design, Diagram, Audit and Dashboards
- Local models, or your own API keys
- 7-day free trial · no card up front
Every model call goes straight from your machine to whichever provider you configure. We never see your code, and there is no usage bill to read.
FAQ
Questions, answered
Can AI generate an architecture diagram from source code?
Yes. Point Lucid Train at a repository and the agent surveys it, then emits a semantic graph of components, tiers and the edges between them. It never places boxes itself: the ELK layout engine computes positions and orthogonal edge routing, which is why the output cannot come out overlapping the way LLM-placed coordinates do. It reads the code on your machine, so nothing is uploaded.
Is there an offline architecture diagram tool?
Lucid Train runs fully offline against local Ollama models. No API key, no telemetry, and your code never leaves the machine. Credentials live in the OS keychain. This is the main practical difference from browser-based diagram tools, which all require sending your architecture to a server.
How much does it cost?
The CLI is free and open source. The desktop app is $3/month, or $30/year, which works out at $2.50/month. There is a 7-day trial. Model usage is separate and never billed by us: run local models at no cost, or bring your own API key and pay the provider directly.
What diagram formats does it export?
Diagrams are Excalidraw scenes, so they export to SVG and PNG and stay editable in any Excalidraw client. One click also turns the canvas into a markdown specification listing every component, every relationship and the narrator's notes.
Desktop app or CLI: what is the difference?
Same agent core. The desktop app adds Code, Diagram, Audit, Dashboards and Connectors in one window with a command palette and themes; the CLI is a single 6 MB Rust binary with a keyboard-first terminal UI. Use either, or both.
Which models does it support?
Any OpenAI-compatible endpoint: local Ollama, OpenRouter (GPT, Gemini, Grok, DeepSeek, Qwen, GLM, Kimi), or hosted cloud models. Different roles can run different models, so a fast model can plan while a stronger one writes the diff. Context windows are tracked and the conversation compacts before overflow.
Install
Design the system. Then let it build itself
One native app for macOS and Windows, or one 6 MB Rust binary for your terminal.
CLI$curl -fsSL https://lucidtrain.com/install.sh | sh