Each animation below is a real moment in the product — a line typed, a diff reviewed, a package exported — shown at full size. Every one links to the page that explains it. Animations play only while they are on screen.
What the editor does for you because your process is text — and no canvas tool can copy.
Type a line, get a box. There is no render button, no layout pass to wait on, and nothing to drag back into place.
Read the page →The selected lines move into a new process and a single .do line takes their place, already linked.
Read the page →A badge counts the callers. Open it to list every .go and .do site with its document and line, and click straight through.
Read the page →Change a document name in the top bar and every .go and .do that points at it is rewritten across all your documents.
Read the page →Type a dot for the command menu, .do for your documents, .back for the steps in this document, an actor’s first letters in a sequence diagram.
Read the page →A dot opens it, a prefix filters it, Enter inserts a scaffold. It sits under your line, flips above it when the window ends, and never covers what you are typing.
Read the page →Fold a decision down to one line. Line numbers keep their real values and the diagram still shows the whole flow.
Read the page →Reference a document you have not written yet and Runic flags the line, then offers to create it right there.
Read the page →Drag a step and its children travel with it. Tab and Shift+Tab indent a whole selection at once.
Read the page →A .note line pins to the step above it and renders as a small comment icon. Click to read, click away to hide.
Read the page →Flip between top-down and left-to-right, pack branches tight, caption arrows, and let .back arrows route themselves around the boxes.
Read the page →The editor, the command menu, the diagram, the chat and the dialogs, key by key.
Read the page →An agent that edits documents the way a careful colleague would: proposals wait as diffs, every change is a checkpoint — and, from one brief, a whole linked system, with the cost of every call on the ledger.
A chat panel scoped to the open document that answers with an exact diff, asks before it assumes, and writes nothing until you click Apply.
Read the page →Each change the agent applies stores the document as it was; ↩ Restore returns to that state at any later point.
Read the page →While a proposal is pending, the flow pane draws the suggested document with the new steps glowing green.
Read the page →Select lines, press Copy to chat, and the agent may rewrite only those lines — everything outside them is carried over verbatim.
Read the page →Instant structural checks with clickable line numbers — thirteen codes, from one-branch decisions to guard fall-through — followed by an AI analyst reading the process for what rules cannot see.
Read the page →Chat history, archived summaries and checkpoints are stored per account, so the agent remembers across sessions and devices.
Read the page →Create with AI plans the product from your brief, writes every process and sequence diagram it needs, links them, checks and repairs each one, then asks what the brief left open.
Read the page →A queue of questions, each bound to one document, asked one at a time; every answer comes back as a diff on the document the question belongs to, and the next question waits until you decide.
Read the page →Shape the idea here, then hand your AI tools a complete spec instead of a guess.
Press Build, tick the documents, answer up to eight questions, and download a repo-ready folder: CLAUDE.md under 150 lines, a plan-mode KICKOFF.md, docs/spec.md with a roles by permissions matrix, a typed data model, contracts derived from your sequence diagrams and acceptance criteria that cite branch ids, vertical-slice milestones, a 13-section design brief, path-scoped DSL rules, hooks when the test command is known, and every ticked flow as a .runic file, checked against ten quality tells before you download.
Read the page →Before writing the spec, Runic reads your ticked documents, the brief and the plan, and asks what a developer would still need to know: the stack and its commands, roles and what each may do, the data behind each step, failures, external systems, numbers for the targets, how users sign in, and the design tone. Answers go into the spec; what you skip becomes an open question.
Read the page →KICKOFF.md is a prompt under 40 lines: read the files in order, run Milestone 0 if the stack is unknown, ask for the sign-in method if nothing states it, then for each milestone plan, branch, write failing tests named by id, implement, verify, review with a fresh subagent, commit, and ask before the next. CLAUDE.md keeps it honest afterwards; .claude/ holds what a request cannot.
Read the page →F1.D2.evet is the Evet branch of the second decision of the first flow; F1.X1 its first failure; S1.M3 the third message of the first sequence diagram; US-1.AC1 a criterion, SC-1 a screen, NF-1 a target, M1 a milestone. Runic assigns them from the parsed text, never from the model, so tests, commits, criteria, transitions and artboards all join on the same names.
Read the page →The done screen runs ten checks on the rendered files, not on the spec object: branch coverage, message coverage, one When per criterion, no placeholders, field closure, role closure, numbers in NFRs, milestone closure, verbatim consistency and budgets. Each shows a tick or a cross with a detail naming the ids at fault, ten points each, and none of them blocks the download.
Read the page →docs/design-brief.md carries the product and the verbatim brief, the personas, tone, platform and breakpoints, one journey per flow walking its happy path with branch ids, a screen inventory with empty, loading, error and success states each, the components reused across screens, copy rules, WCAG 2.2 AA numbers, what to design first, what is out of scope, every flow embedded with the legend, the integration touchpoints per sequence, and the handoff. Paste it into Claude Design alone, or run /design on it inside Claude Code.
Read the page →export_build_package returns the same files and the same ten tells the app shows, as JSON, with a read-only token or a claude.ai connection. The build-spec prompt makes Claude Code run the interview itself, write the typed spec, call the tool with only: "lint" until nothing fails, fetch the files whole with paths, and write them into the repo, including .claude/. Runic's server calls no model.
Read the page →Everything a diagram tool gives you — links, embeds, exports, history, teams — reading from one text.
Two clicks mint a /s/ link anyone can open. It renders what the document says right now, and Revoke kills it on the spot.
Read the page →Paste one iframe into your wiki or docs page. The diagram inside re-fetches the document every minute.
Read the page →Every export reads from the same text as the editor, dressed in the style you pick, and named after the document.
Read the page →Runic records the document as you save it, keeps the 50 newest automatic snapshots, and never drops the ones you name.
Read the page →Create a team, invite people by email or open link, and every document created inside it is theirs to read or edit.
Read the page →Map draws your whole workspace as a graph: processes as circles, sequence diagrams as rounded squares, .do and .go lines as arrows, missing targets as hollow red ghosts.
Read the page →Ten single-word commands for flows, a handful of arrows for sequences. Indentation carries the structure.
Plain lines are steps. A dot starts a command. Two spaces of indent carry the structure, and the flowchart redraws from the text on every keystroke.
Read the page →One interaction per line. Write the arrow you mean, and the actors, lifelines, loops and alternatives draw themselves.
Read the page →Sign-in, MCP for Claude, and the privacy choices that shaped how Runic is built.
Runic runs an MCP server with 18 tools, six prompts, resources and a change stream; connect from claude.ai with a click, from Claude Code with one command, read-only or read and write.
Read the page →One account, three ways in: a password, Google or GitHub, linked by verified email and managed from Settings → Connected accounts.
Read the page →Runic sets three first-party cookies and nothing else, loads nothing from third parties, keeps AI keys on the server and gives you export and deletion under Settings → Your data.
Read the page →Looking for the full text of a feature? Every feature, one page each →