Choose the input you already have
FlowGraph's free tools are not separate miniature products. Each one is a short front door into the same living graph. The practical question is therefore not “which diagram maker should I learn?” but “what material do I already have?” Start with an IFC model when the building is the source, Mermaid text when an AI or repository already produced the flowchart, or a plain-English description when the process is still in your head. The dedicated tool doors open without an account so you can view, convert, or generate first; sign-in appears when you save or continue with gated AI. The result stays on your device unless you deliberately use a sharing or cloud feature.
That shared destination matters. A normal converter produces a picture and ends. FlowGraph produces typed cards and connections that can be rearranged, queried, assigned, reviewed, and shown through multiple synchronized lenses. A decision visible in the canvas is the same decision in the outline, timeline, board, and executive view. There is no export-and-rebuild step between them, so one correction updates every view.
Mermaid to editable flowchart
Use the Mermaid converter when ChatGPT, Claude, a Markdown file, or project documentation already contains a flowchart. Paste supported graph TD, graph LR, or flowchart syntax and FlowGraph converts its nodes and directed edges into native cards and connections. Common shapes and pipe-form edge labels are supported. Subgraphs, styling directives, class definitions, sequence diagrams, and other Mermaid diagram families are not yet full-fidelity imports, so the page says “Mermaid flowchart” rather than implying universal Mermaid compatibility.
The conversion runs on the device. Your diagram text is not uploaded merely to parse it. After conversion, rename a card, reconnect a branch, add an owner, or move the same graph into another lens. This is currently the real cross-tool import path. FlowGraph does not yet ship direct Lucidchart or Miro import adapters, and these pages do not claim otherwise.
AI diagram generator from text
Use the AI diagram generator when the process is easier to describe than to draw. State the beginning, the desired outcome, and the decisions in between. “Map customer onboarding from sign-up to first value, with a branch when identity verification fails” gives the generator clearer structure than “make an onboarding diagram.” The result is a draft graph whose changes remain reviewable; it is not a flattened illustration.
The free generator door includes a hosted path subject to fair-use limits, and you can bring your own model key after entering the app for more control. AI output should be treated as a proposal. Check names, order, branches, and assumptions before using the graph as shared truth. FlowGraph's governing boundary exists precisely because fluent output is not the same thing as verified work.
IFC viewer with no model upload
Use the IFC viewer when the source is a building model. Standard IFC files are parsed and rendered in the browser, so the model bytes do not need to be uploaded to FlowGraph's servers. You can inspect the spatial tree, classes, elements, and available property sets, and view geometry when the browser supports WebGL. The local no-upload path is for IFC; native Revit, Navisworks, and AutoCAD formats require translation rather than pretending they are IFC.
The viewer becomes more useful when a model element is connected to the work around it. A door can link to its submittal, inspection, field issue, and handover evidence. An RFI can point to the beam it questions. The model then stops being a file that is occasionally opened and becomes one lens over the same project graph. The construction hub collects the viewer with process templates designed for that job.
What “free” and “local-first” mean here
Free has two entry boundaries. A dedicated tool or template link opens account-free so you can see the result before the save gate. The ordinary browser app and the pip-served app require a free account at first run so identity and entitlements are explicit; afterward the local session can continue offline, and your local vault is not held hostage by an expired network session. Pro is for capabilities such as live command-line execution and other working-tier features, not for installing the local package or opening the basic canvas.
Local-first means content remains on the device by default. It does not mean the product can never make a network request: sign-in, deliberate sharing, hosted AI, sync, and opt-in cloud translation are network features. Each tool describes its own boundary. For example, the Mermaid parser and IFC parser can work on-device, while hosted AI generation necessarily calls a model service. The honest distinction is more useful than a blanket “offline” label.
Prefer a known pattern?
If you do not have a file or a description yet, start from the template gallery. Its previews cover ordinary flowcharts, decisions, roadmaps, mind maps, knowledge graphs, organization charts, and construction processes. A template opens account-free into the same graph model as every tool above, so it can be explored and extended before the save gate rather than redrawn. If you want the product sequence before choosing, the five-minute guide explains the core objects, review loop, and lenses in plain language.