← Back to Lab

Building with AI Assistants

I've been experimenting with OpenClaw, an AI assistant framework that runs locally and integrates with real-world tools and services. Instead of treating AI as just a chatbot, OpenClaw can execute commands, manage infrastructure, and automate workflows across Google Workspace, GitHub, and more.

What I Built

A personal AI assistant that can:

  • Manage Google Workspace: Send emails, check calendar availability, create contacts, and search Drive—all via the gog CLI tool
  • Automate GitHub workflows: Clone repos, create branches, make code changes, and submit pull requests on my behalf
  • Update this website: My assistant wrote the About page content by reading my resume from Google Drive and creating a PR with professionally-written copy
  • Search and fetch web content: Research job postings, fetch documentation, and summarize findings

Why This Matters

Traditional automation requires writing scripts for every task. With an AI assistant that understands natural language and has access to CLI tools, I can say "update my About page based on my resume" and it handles:

  1. Reading the resume from Google Drive
  2. Analyzing the content
  3. Cloning the repo
  4. Making appropriate edits to multiple files
  5. Creating a PR with a detailed description

All in one conversational exchange.

Technical Architecture

Core Stack:

  • OpenClaw (AI assistant framework)
  • Claude Sonnet 4 (primary model)
  • Grok 4 (secondary agent for specific tasks)

Integrations:

  • Google Workspace via gog CLI (Gmail, Calendar, Drive, Contacts)
  • GitHub via gh CLI (repos, PRs, issues, actions)
  • Web search and content fetching

Security Model:

  • Runs entirely locally (no third-party API proxies)
  • OAuth-based authentication for Google services
  • GitHub token with repo-scoped permissions
  • All credentials stored locally, never in code

Real Examples

Example 1: Calendar + Email Workflow

I asked: "What time slots do I have available tomorrow?"

The assistant:

  • Queried my shared Google Calendar
  • Parsed the events
  • Sent a formatted email with openings

Example 2: Site Updates

I asked: "Update my About page with content from my resume in Google Drive"

The assistant:

  • Downloaded my resume (Word doc) from Drive
  • Extracted and analyzed the content
  • Cloned the personal-site repo
  • Rewrote both the About and Home pages with professional copy
  • Created a PR with detailed commit messages

I reviewed, merged, and the changes went live.

Example 3: Job Research

I asked: "Find the PGA Tour Superstore senior engineering manager opening"

The assistant:

  • Searched the web for the job posting
  • Scraped the listing from a job board
  • Extracted key requirements (Salesforce Commerce Cloud, CI/CD, Kubernetes)
  • Emailed the details for review

All without me leaving the chat.

Why I'm Excited About This

As someone who's built internal developer platforms at scale, I'm fascinated by how AI can be the next evolution of self-service tooling. Instead of building UIs and APIs for every workflow, an AI agent can:

  • Understand intent from natural language
  • Orchestrate multiple tools and services
  • Handle edge cases conversationally
  • Learn from feedback in real-time

It's like having a platform engineer on call 24/7.

What's Next

I'm exploring:

  • Multi-agent workflows (separate agents for different contexts)
  • Cron-based automation (scheduled check-ins, reminders)
  • Memory and context management (how agents remember past interactions)
  • Security hardening for production use

This is early-stage experimentation, but it's a glimpse of how AI assistants could change how we interact with infrastructure and tooling.

Tech Stack: OpenClaw • Claude Sonnet 4 • Grok 4 • gog CLI • gh CLI

Languages: TypeScript • Go • Bash

Platforms: Google Workspace • GitHub • Web APIs