Your team makes technical decisions on gut feel. It does not have to be that way.
ChatGPT gives you a paragraph. grainulation gives you typed claims with conflicts highlighted. Research, challenge, compile — then ship a decision brief your team can audit. Research that compiles.
Everything in grainulation starts with a claim — a single typed statement with an evidence grade. Claims have types (factual, risk, estimate, constraint, recommendation) and tiers (from "someone said it" to "measured in production"). The compiler validates them. The tools help you collect them.
grainulation is an ecosystem of CLI tools for structured technical research. Start with wheat — you will discover the others as you need them. Ask a question ("Should we migrate to GraphQL?"), research from multiple angles, challenge your findings, and compile into a decision brief.
The compiler is JavaScript code, not an LLM call. Same claims in, same result out, every time. It catches contradictions, flags weak evidence, and blocks output until issues are resolved. Your decision brief has a git audit trail showing how every claim was collected and challenged.
You need Claude Code, Node.js 20+, and npx. No accounts, no cloud services, no API keys for the core tools. Zero npm dependencies across the entire ecosystem — built on Node.js built-ins only. Everything runs locally, everything is plain JSON and HTML.
"We should probably use GraphQL because it's more modern." One paragraph, no sources, no way to challenge it.
14 typed claims across 4 topics. 3 risks flagged. 1 contradiction caught. Recommendation with evidence grade. Full git audit trail.
From question to shipped decision
Start with wheat. Add tools only when you need them. Most sprints need just the core engine.
Three commands. One decision.
No install. No signup. Just npx and a question.
Initialize a sprint
Start with the question you need answered. wheat creates a claims file and config.
$ npx @grainulation/wheat initResearch and challenge
Gather evidence, then stress-test it. Every finding becomes a typed, graded claim.
wheat> /research "topic"
wheat> /challenge r001Compile and ship
The compiler resolves conflicts, checks evidence, and produces a decision brief.
wheat> /briefQuestion to decision in 10 minutes
A team needs to decide whether to migrate from REST to GraphQL. Here is the full sprint, start to finish.
Eight tools. Each does one thing well.
Start with wheat. Everything else is optional. Add tools as your research practice grows.
Built on four principles
The decisions behind the tools.
Satisficing over maximizing
You will never have perfect information. The goal is enough evidence to make a defensible decision, not a perfect one.
Claims over opinions
Every finding is a typed claim with an evidence tier and provenance. "I think" becomes a traceable claim ID with a grade.
Adversarial pressure
Comfortable agreement is the enemy of good decisions. The system forces you to challenge, witness, and stress-test every claim.
Zero dependencies
Node.js built-ins only. Zero npm dependencies across all eight packages. No supply chain anxiety. No left-pad. No API keys for the compiler. Every tool ships what it needs and nothing more.
Common questions
Yes. Node.js 20+ is required. That is the only system dependency. Everything runs via npx with zero npm install required — packages are fetched and cached automatically.
grainulation is language-agnostic. It researches decisions about any technology — databases, APIs, infrastructure, architecture. The tools run locally, but the research can be about anything.
RFCs and ADRs document a decision after it is made. grainulation captures the research process — evidence gathering, adversarial testing, and conflict resolution — and validates it through a compiler. The output brief can serve as your ADR, with a full git audit trail showing how every claim was collected, challenged, and resolved.
No. Start with wheat alone — it handles the full research-to-decision workflow. Add other tools only when you need specific capabilities: farmer for AI permissions, mill for export, silo for reusable knowledge, and so on.