Positioning
Where cix sits in the developer tools landscape, in straightforward language.
The category
cix is project-context infrastructure for AI coding assistants.
It is not an AI coding assistant. It is not a code search tool. It is not an IDE extension. It is the layer between your AI assistant and your codebase that gives the assistant a real, queryable understanding of the project it is working inside.
This is a new category, but it is not a hypothetical one. As AI coding assistants become a central part of how teams write software, the gap between what the assistant knows and what your project actually contains becomes a first-order productivity question. cix is the infrastructure that closes that gap.
What cix is built around
Three convictions shape every product decision we make:
1. Grounding beats fine-tuning. Teaching a model to know your codebase by training is expensive, brittle, and opaque. Grounding the model in a queryable index is cheap, robust, and transparent. The same outcome, with much better economics and visibility.
2. Honest "I don't know" is a feature, not a failure. Every coding tool that fabricates plausible-looking nonsense earns a reputation for unreliability. cix is designed to fail openly — to say "no parser," "partial schema," "low confidence" — rather than make something up. That honesty is what makes the confident answers trustworthy.
3. Structural enforcement matters more than structural description. A document describing conventions decays. A system that enforces conventions at write time holds. The difference at six months is the difference between a clean codebase and a drifted one.
Where cix sits relative to the alternatives
A short version (the long version lives on the Comparison page):
| If you are using | What cix does for you |
|---|---|
| Grep / ripgrep alone | Replaces grep's use by the AI assistant with a structured query layer; you keep grep for human work |
| An LSP in your editor | Adds the same kind of structured understanding for the AI assistant; you keep the LSP for editor work |
| Vector search / RAG over the codebase | Likely replaces it for most code-shaped questions, with stronger guarantees |
| A CLAUDE.md or AGENTS.md instruction file | Complements it; cix grounds the factual layer, instruction files describe intent |
| Fine-tuning on your codebase | Likely replaces it for most teams; grounding is cheaper and more transparent |
| Nothing | Addresses the structural class of AI failure modes that compounds silently |
Who cix is for
Mid-size and large engineering teams using AI coding assistants on real codebases. The teams seeing the most value are the ones spending real reviewer time on small mechanical issues today.
Solo developers maintaining a handful of medium-sized projects who want the assistant to stop duplicating their own helpers and dropping files in the wrong place.
Technical evaluators comparing AI tooling for adoption. The case-study and benchmark pages on this site are written specifically for that audience.
Who cix is not for, today
Codebases primarily in unsupported languages. Lua is the canonical example. The system is honest about this; the value is small until the parser lands.
Greenfield projects without established conventions. The system can still help, but you'll be authoring conventions from scratch rather than benefiting from inferred ones.
Teams that don't use AI coding assistants. This is obvious but worth saying. cix is infrastructure for AI assistants. Without one, the value isn't there.
Single-file scripts and one-off projects. Below ~50 source files, you can probably hold the project in your head; the index is overkill.
What we charge for
This is a public docs site so we are not prescriptive about pricing here. The general framing:
- Parsing is local; metadata indexing is cloud. cix reads your code on your machine and ships only structural metadata (symbols, signatures, call graphs) to cix-cloud. Your source never crosses the wire.
- Per-user index, by design. No two users share an index — even within Team accounts. The trust pillar is data isolation, not absence of cloud.
- Adoption is incremental. A single developer can install and use cix on a single project. There is no need to commit at the team or organization level before evaluating.
For the exact commercial details, see the project README or contact us directly.
How we differentiate
Honest behavior. Most tools in this space optimize for confident-looking output. We optimize for grounded output, and we surface the gaps explicitly. This shows up everywhere: in the parse-error reporting, in the partial-schema flags, in the confidence tiers on impact analysis. The honest-behavior commitment is the single most important thing about cix.
Multi-assistant by design. cix works with Claude Code, Codex, and Gemini CLI through the same integration standard. We don't pick a winner. Your team's tooling choice is yours; the index is portable across whatever you use.
Local-first by design. Source code stays on your machine — only structural metadata reaches cix-cloud. Each user gets a private, isolated index. The trust pillar is data isolation, not the absence of cloud.
Convention enforcement, not just convention description. The convention layer can block writes that violate project rules on supported clients. This is meaningfully different from a markdown file that describes what should happen.
Shipped with sensible defaults. Convention profiles for ten common stacks ship out of the box. You start with something reasonable and tune. You don't start from scratch.
What we are not
- A coding assistant. We don't generate code. We give your assistant the context to generate it well.
- A code review tool. We surface findings; we don't replace human review.
- A test runner. We tell you what depends on what; running the tests is your problem.
- A migration tool. We parse migrations into a schema view; we don't write migrations for you.
- A vendor lock-in. Drop us tomorrow if you want; the index is local, the conventions are committed to your repo, and the integration standard is open.
A one-paragraph summary
cix is project-context infrastructure for AI coding assistants. It builds a structured, queryable index of your codebase and exposes it to your assistant through an open integration standard. The result is an assistant that searches before it writes, respects your project's structure, references real schema instead of inventing it, and tells you when it doesn't know something. It works with the major AI coding tools, parses locally so your source never leaves your machine, ships with sensible defaults for the common stacks, and is honest about what it can and cannot do. That is the entire pitch.
Related
- Overview — longer pitch.
- Comparison — vs. the alternatives.
- Limitations — where the system underperforms.
- Roadmap — what we're working on next.