Getting started.

Sign in

Open Corenel with Get started. The workspace, Crew, Workflows and the console ask you to sign in first; Activity, Files and the recorded demo run open without an account. The account is free and asks for no card.

Add a model

Corenel does not supply a model. Until you add one, the model picker says No models yet and offers Add an API key, which opens Settings on the Models tab. Settings lives in the account menu behind your avatar. You have three ways in:

Built-in

Pick a provider from the list and paste its API key. The key is stored encrypted with your account, and runs on that provider go through our gateway, billed to you by the provider.

Custom, your endpoints

Choose Add custom provider and give it a name, the base URL of any OpenAI-compatible server (Ollama, LM Studio or vLLM on your machine, or another vendor’s compatible API), its models, and an API key if it needs one. Your browser calls it directly, so that key never reaches us.

Claude Code on your machine

Start the sidecar with --acp claude-code and Claude Code appears in the model picker under On your machine. See Running the sidecar.

Give the agent a folder

Without a sidecar the agent works in a workspace stored in your browser. To let it read and write real files, install and pair the sidecar on the machine that has them, and pick the folder in the location bar at the top of the workspace. The whole walkthrough is in Running the sidecar.

Chat with the agent

The workspace is a chat with the agent beside your files. Pick a model in the header, type, and send with Cmd/Ctrl+Enter. The + button attaches a document, and you can paste a screenshot straight into the composer.

The sliders button in the composer opens this chat’s options. Its Policy row chooses what the agent may do on its own: standard (reads run, writes and commands ask you), read-only (writes and commands are refused), or autonomous (everything runs without asking, except changes to Corenel’s own configuration). The same pane can switch a tool source off for one chat.

Projects and documents

The Projects list in the workspace rail holds your projects; its + button starts one. Ask for a document in a project’s chat and the draft appears in the document panel beside it. Export from the panel header: a document as PDF, HTML or DOCX, a spreadsheet as XLSX, a slide deck as PPTX, an image as PNG. Export happens in your browser, with no upload.

Workflows, Activity, Files and the console

Workflows

A canvas for fixed steps (API calls, mappers, branches, loops) alongside prompt, agent and guardrail nodes where a model judges.

Activity

Two tabs. Runs replays recorded runs with a timeline scrubber; up to the hundred most recent are kept, and a pinned one is kept until you delete it. Live shows the sessions a paired sidecar is holding open.

Files

Browses the folder the agent is working in and is where you review an edit it proposes. It writes nothing except an edit you accept.

Console

Opens a command console over whatever page you are on (Ctrl+`). Type help to list its commands.