Local-first coding agents

How to Use a Local AI Coding Assistant in VS Code

Set up a local or BYOK coding assistant in VS Code with Truss, then use read-only chat, planning, agent tools, and model profiles without leaving your editor.

local AI coding assistant VS CodeOllama VS Code coding agentVS Code coding agentBYOK VS Code extensionlocal LLM VS Code

Keep the agent inside the editor you already use

A VS Code coding assistant should add useful project context without forcing you into a separate editor or hosted workspace. Truss runs as an extension surface over the same provider-neutral runtime used by its other clients.

That means your model can be local or BYOK, while the interaction stays close to the open file, active workspace, Git changes, and editor workflow.

Start in Chat or Plan before Agent

Use Chat for questions like ‘what does this module do?’ or ‘where is this value created?’ It can inspect safe workspace context without making a change. Use Plan when you want a read-only investigation followed by proposed implementation steps.

When you are ready to make a change, switch to Agent and choose the tool permission policy intentionally. This sequence creates a more reliable review loop than asking every conversation to be autonomous.

Configure the model once, then choose it deliberately

In Truss settings, connect a local endpoint or create a named cloud provider account. Use Discover models where the provider supports it, verify the exact model, and run a connection test before assuming an agent failure is a code problem.

The extension exposes the active model and context information so you can check the foundation of a response. If a small local model stalls on tool use, change the model or narrow the task rather than repeatedly sending the same prompt.

Use VS Code as a review surface

The best agent workflow does not replace your editor’s normal strengths. Keep source control visible, inspect generated diffs, and use the regular terminal for commands that deserve your full attention.

Truss is meant to fit that loop: ask, inspect context, review proposed work, allow only the tools you want, and validate the result using the same editor and Git tooling you already trust.

Next steps

Put it to work in your own workspace.

Pick the surface that fits your day, then connect a model you control.

FAQ

Questions developers ask before switching.

Can Truss for VS Code use Ollama?

Yes. Configure a reachable local endpoint, select an installed model, and use it from the extension.

How do I install the VS Code extension?

Download the signed VSIX from the current Truss GitHub Release, then use Extensions: Install from VSIX in VS Code.