LLM Cheburnexus

Your AI assistant burns tokens reading files. This makes it stop.

Cheburnexus indexes your project at compiler level — the same understanding a compiler has, not text search — and serves your LLM exactly the code it needs. TypeScript/JavaScript, Python, Java, C#.

Works across the whole VS Code family: VS Code, Cursor, Windsurf, VSCodium.

Why it makes your AI better

Without Cheburnexus
  you --> LLM --> read file --> grep --> read file --> read file --> grep ...
              |__ context filling up with raw text __|
                                                     --> answer

With Cheburnexus
  you --> LLM --> Cheburnexus --> structured facts --> answer
                  (compiler-level index)

Wasted context is not just cost — it is worse answers. A model that spent its window on boilerplate has less room left to think about your actual problem.

Cheburnexus hands it a structured project instead: classes, call graphs, dependencies, cycles, entry points — resolved by a real compiler front-end, not guessed from patterns. Language models are dramatically better at reasoning over structure than over raw file dumps.

Every response reports how many tokens it saved you. The optional statusline keeps a running total.

Using it

Just ask normally — but point the model at Cheburnexus. It works for far more than search:

• Finding code — "Use Cheburnexus to find where user authentication is handled."
• Before editing — "Check with Cheburnexus who calls PaymentService.refund before you change it." The model sees the blast radius instead of discovering it after breaking something.
• Understanding a new project — "Give me a Cheburnexus summary of this codebase, then explain the main entry points."
• Reviewing — "Ask Cheburnexus for cycles and findings, and tell me what's worth fixing."
• Refactoring — "Use Cheburnexus to map dependencies of the billing module before we split it."
• Tracing a bug — "Trace the dependency chain from the API handler down to the database layer via Cheburnexus."

After you change code, ask the assistant to reanalyze so the index reflects the latest source.

If the model forgets

Language models are probabilistic — yours may drift back to reading files by hand, especially in a long session. That is normal, and you fix it in one line:

"Use Cheburnexus for this, not file reads."

Worth saying once at the start of a session, and again whenever you notice it grepping.

To make it stick permanently, put the instruction in your assistant's project rules file — it is then loaded into every session automatically:

• Claude Code — CLAUDE.md in your project root
• Codex / Cursor — AGENTS.md in your project root

Always use Cheburnexus tools to explore this codebase before reading files directly.
Prefer get_summary, find_class and who_calls over reading or grepping files.

What it gives the model

Project overview, class and method lookup, who-calls / usages, dependency graphs and chains, namespaces and layers, dependency cycles, code findings, metrics and outliers, focused file content, re-analysis on demand.

Crucially, it answers precisely or stays silent — it reports facts it can prove, never guesses. If a signal is unknown, it says unknown rather than inventing an answer for your model to trust.

Install

1. Install the extension.
2. On first run it downloads the analysis engine — one time, verified by checksum. It is not bundled in the extension, so the download is not instant; progress is shown in the status bar.
3. Open your project. The MCP server registers automatically; there is nothing to configure.
4. Ask your assistant something from Using it above.

The Cheburnexus panel in the activity bar shows your plan, engine status, and token savings.

Free plan works immediately, with no key. Paid plans: place your claims.json license file in the path shown in the status panel, then restart the IDE.

Support & feedback

Questions, bugs, feature requests: t.me/cheburnexus.

If something misbehaves, do not just say "it doesn't work" — the plugin can gather the evidence for you. Run Cheburnexus: Collect Problem Report from the command palette (or click Report a Problem in the Cheburnexus panel). It assembles the analysis log, versions and your scope settings into one file.

That file is opened locally for you to review and is never uploaded. It contains paths and folder names from your project — read it, delete anything you would rather not share, then send it.

Privacy

All analysis runs locally on your machine. Your code is never uploaded anywhere. See PRIVACY.md for details.

Contributing

Building, packaging, and the extension's internals are documented in CONTRIBUTING.md.