I Built My Own Jarvis with Claude Code (No Code Setup)

Summary

The video demonstrates how to build a personal AI assistant using OpenJarvis and Claude Code. Setup and configuration are handled through plain-English instructions rather than manually writing code.

How the system works

• OpenJarvis is the assistant framework. It runs on the user’s computer and can interact with files and connected apps, depending on the permissions granted.
• Claude Code acts as the setup and troubleshooting tool. The presenter uses it in a terminal to install OpenJarvis, configure models, and fix errors.
• An AI model provides the assistant’s “brain,” while access to local files and apps lets it perform practical tasks.

Setup guide

1. Install Claude Code, create an empty project folder, open it in a terminal, and ask Claude Code to install and configure OpenJarvis.
2. Create an API key through NVIDIA’s model platform and provide it during setup. The video uses NVIDIA models, including Nemotron 3 Super, and mentions other models as alternatives.
3. If setup produces an error, paste it into Claude Code and ask it to fix the issue.
4. Open the assistant in a browser and try tasks such as summarizing a folder, finding a document, building a basic website, or researching and drafting a report.

Models and voice options

• NVIDIA models: Presented as the free, cloud-based starting option.
• Ollama: Runs a compatible model locally, allowing prompts to stay on the computer and enabling offline use after setup. Local models may respond more slowly, particularly on computers without a dedicated GPU.
• Claude API: An optional paid model for more demanding reasoning, writing, analysis, or coding. API charges are separate from a Claude.ai subscription and depend on usage.
• Local speech tools: Add speech-to-text and text-to-speech so users can speak through a microphone and hear spoken replies.

Remote access through Slack

The video shows how to connect OpenJarvis to a Slack workspace so it can be messaged from a phone. Setup uses a Slack app manifest and two tokens: a bot OAuth token and an app-level token with the connections:write scope.

Messages sent through Slack are handled by Jarvis on the user’s computer. The computer must remain on and connected to the internet for this to work.

Key cautions

• Start with a test folder and limited permissions. If enabled, Jarvis may be able to read, create, or modify files and run commands.
• Review the permissions requested by connected accounts and integrations.
• Keep API keys and Slack tokens private.
• Monitor billing when using paid APIs. Frequent automated requests or processing large files can increase costs.

Main speaker and sources

The video features a single presenter and narrator from the Outskill channel; no individual speaker is identified in the subtitles.

Tools and services discussed: OpenJarvis, Claude Code by Anthropic, NVIDIA models, Ollama, and Slack.