What is Superpositional?
A code intelligence platform that indexes your repositories and gives you answers, reviews, and drift detection across the full engineering lifecycle.
Superpositional is a code intelligence platform built by State Vector Co. It indexes your repositories into a hybrid vector + graph store, then uses LLMs to answer questions, review PRs, generate documentation, detect staleness, and investigate issues.
The tagline: Ship fast. Break less. Know all.
What it does
Most AI tools help you write code. Deployment, security, QA, operations, incidents? Undersupported. Superpositional is the intelligence layer across the full engineering lifecycle — not just the editor.
It doesn't just observe what happened. It traces why — following the causal chain from symptom to root cause through the code graph. A doc is stale because a function signature changed, which broke a contract, which affects these consumers. Not flags. Cause and effect.
Who it's for
Small teams managing production systems. You don't have dedicated platform, SRE, or QA teams. Superpositional gives you that operational leverage — rapid incident root-cause analysis, proactive PRs from customer-raised issues, enforced coding standards that compound over time.
Growing teams where nobody knows what depends on what. Cross-repo visibility, drift detection, architecture health. When the codebase outgrows any one person's mental model, you need a system that holds the map.
Leadership that needs answers in plain English. Release readiness, risk exposure, architectural health — queryable without reading code. You ask questions; you get answers.
Core capabilities
Repository indexing. Superpositional clones your repos, parses them with tree-sitter, generates embeddings, and stores everything in a hybrid vector + graph store. Entities, relationships, and change history — all queryable.
Natural-language querying. Ask questions about your codebase in plain English. Architecture, dependencies, security posture, change history. The chat interface works for developers and non-developers alike.
PR reviews. Every pull request gets checked against the system model. Architectural drift, boundary violations, dependency breakage — caught before merge, not after deploy.
Document freshness detection. Structured claims are extracted from your docs and evaluated against code changes. When code drifts from what the docs say, you know — without burning LLM tokens on every check.
MCP integration. Superpositional is MCP-native. Your AI coding tools (Cursor, Windsurf, VS Code) get structured entities, relationships, and architectural context instead of raw file dumps. Better context in, better code out.
How it's different
You pay for operations, not seats. Usage-based pricing means your team scales without a per-seat tax.
AI-generated code accrues debt faster than you can track. Standards drift, docs go stale, architecture erodes. Superpositional detects that drift continuously, so the codebase stays liveable.
Other tools observe PRs in isolation. Superpositional maintains a persistent model of your entire codebase and traces causality through it. That's the difference between "this line looks wrong" and "this change breaks three downstream consumers and makes two docs stale."