The cleanup crew for your coding agent

codalotl is the cleanup crew for your coding agent. It's a local CLI for Go that turns messy or machine‑generated code into production‑ready changes—fixing documentation, deduplicating repeated logic, organizing package layout, and normalizing names. Under the hood it uses fast AST‑based analysis plus targeted model calls, and outputs small, reviewable diffs you can commit with confidence.

Agents try to build features. codalotl refines code you already have.

• Scope: focused on docs, duplication, layout, and naming—not inventing business logic.
• Structure-aware: uses the Go AST to make precise, cross-file edits that keep intent intact.
• Reviewable: produces tight diffs instead of sprawling rewrites.
• Tool-friendly: runs alongside your formatter, linter, and CI—no lock-in, no scaffolding.

Not reliably - especially across a whole package. Agents are great at one-off, explicit edits ("document this function"). The moment you ask for "document everything" or "rename things consistently," they skip items, stop early, or forget parts of your style guide. The more conventions you stack, the more brittle they become.

What codalotl does differently

• Complete coverage. We walk the Go AST to inventory every package/type/func that needs attention, then verify the result - so nothing gets missed.
• All your conventions, enforced together. GoDoc style, idiomatic naming (including acronyms), and file layout are applied consistently rather than "best-effort."
• Behavior-preserving refactors. We find structural duplication, extract helpers, update call sites, and normalize names across files - emitting small, reviewable diffs.
• Reliable process. Changes are decomposed into focused steps with targeted model calls and automatic checks, yielding consistent results run-to-run.

Use your agent to build features. Use codalotl to make the result consistent, fully documented, and ready for PR.

Doing this well is language-specific. Go has unique documentation conventions (GoDoc), idiomatic naming rules (including acronym handling), and well-worn ways to organize packages and files. codalotl walks the AST to understand what’s documented, what’s duplicated, and what should move—so it’s tuned deeply to Go. Other languages may come later, but only when we can match this level of quality.

Yes. At some point in the distant future, when all the features are extensively fleshed out, codalotl may have a small fee (which would be communicated well ahead of time).

No. We love OSS, but codalotl itself isn't open source and we don't plan to open it.

No. You bring your own API keys or run your own LLMs, and requests go directly from your machine to the model provider. We don't sit in the middle.

codalotl works with major providers and OpenAI-compatible “drop-in” APIs. If your provider exposes an OpenAI-style completions/chat endpoint, it likely works. You supply the keys; swap models anytime without changing your workflow.

Yes. Analysis and edits run locally, and your code never touches our servers. Any model calls go straight from your machine to the provider you configured. If you self‑host a model, nothing leaves your network.