sprang

The qualitative leap in codebase comprehension.
Det qualitative Spring — Kierkegaard

---

Sprang is a knowledge graph platform for Windsurf (Cascade / Devin Desktop), Claude Code, and GitHub Copilot that creates total comprehension of codebases, knowledge bases, and document vaults — not just symbol search, but why code exists, who changed it, what it risks, and how it all fits together.

Your AI agent is the intelligence layer. Sprang is the memory. Together they answer "what will break if I change this file?" in a single tool call — and "how does this codebase actually work?" for anyone who just joined the team.

"The System knows everything about being, but nothing about existence."
Kierkegaard's critique of Hegel applies equally to symbol indexers and grep tools.
Sprang bridges the gap: from static facts to living, contextual understanding.

---

The Leap

Det qualitative Spring — the qualitative leap — is Kierkegaard's name for a discontinuous jump in understanding: the kind that cannot be reached by incremental steps, no matter how many you take.

Symbol search finds where things are. Documentation says what they were meant to do. An LLM can explain individual files brilliantly — and still lose the plot at file 50, forget the conversation from yesterday, and have no way to answer the question that matters most: what breaks if I change this, before I break it?

These answers require different infrastructure — one that understands the codebase before your agent starts working, persists that understanding across sessions, and makes the hard questions answerable in a single tool call:

• Why does this file exist? → sprang_why reads git history, PR references, and team annotations
• What breaks if I change it? → sprang_diff_impact runs BFS over the full dependency graph
• How risky is it? → sprang_health surfaces blast radius × coupling × test gap × churn, scored 0–1
• What does this codebase actually do? → /sprang-onboard gives a persona-adaptive guided tour

The leap becomes repeatable. The graph persists. The context accumulates.

---

Not just codebases

The same infrastructure works for knowledge bases: Obsidian vaults, Logseq databases, Dendron workspaces, Foam wikis, Zettelkasten archives, or any folder of markdown. Notes become nodes. Links become edges. Topic clusters emerge. The same Ask Agent panel, the same force-directed graph, the same guided reading order — just pointed at your notes instead of your code.

/sprang-knowledge /path/to/your/obsidian-vault

---

Installation

Note: Windsurf AI and Devin Desktop are the same product — Windsurf was rebranded as Devin Desktop. All instructions, skills, and workflows are identical for both. Both names appear in this README.

Quick install (npm) — the easiest path for every platform

Two commands set up any project for any agent — no clone, no build, no manual file copying:

npm install -g @faviovazquez/sprang
cd my-project
sprang init --platform claude     # or: copilot | windsurf | all

sprang init --platform <agent> does everything in one step:

• writes the MCP config where your agent reads it — .mcp.json (Claude Code), .vscode/mcp.json (Copilot), or .devin/config.json (Windsurf/Devin) — with the absolute path to the bundled MCP server already filled in;
• copies that agent's slash commands, rules, workflows, skills, and merge.py into the project.

Then build the graph and open the dashboard: