CodeLantern Docs
Core Concepts

Core Concepts

The mental model behind CodeLantern: the workflow your team runs, the knowledge base it reads and writes, and how it all fits together.

CodeLantern has three moving parts. Once you have a mental model of how they connect, the rest of the docs falls into place.

A mental model: CodeLantern's knowledge base and workflow skills run in a local or cloud harness, every call routes through the cl-platform layer, which connects to project-management integrations (GitHub, Linear, and Azure DevOps supported, Jira coming soon) and source-control integrations (GitHub and Azure DevOps supported).

The three parts

  • Your knowledge base and workflow skills. The skills (spec, design-solution, implement, review-code, …) and the knowledge base are the CodeLantern layer. They're portable across coding agents and run either locally or in the cloud. Locally, you drive them interactively through the cl-dev plugin. In the cloud, they're triggered by @codelantern-ai mentions on an issue or pull request.

  • The CodeLantern platform. Every action a skill takes flows through CodeLantern, which routes and audits the call and handles talking to your providers on your behalf, so you never manage tokens or provider secrets yourself.

  • Your provider integrations. This is where the work actually lives, split into two roles. Project management (PM) holds your issues and board. GitHub Projects, Linear, and Azure Boards are supported today, with Jira coming soon. Source control (SCM) holds your code and knowledge base, where changes land as branches and pull requests. GitHub and Azure Repos are supported today. CodeLantern is provider-agnostic by design, so new integrations are additive.

Go deeper

On this page