Journal
The Le Bureau notebook.
What we learn building an office for AI agents. Technical decisions, user feedback, behind the scenes. RSS
Drive your cloud desktops from Claude: our MCP server is live
@lebureau/mcp is on npm. Your Le Bureau desktops show up as tools that Claude Desktop, Claude Code or Cursor can call directly — with per-key capability scoping to fence off what each integration is allowed to touch.
Why Your AI Agents Need a Shared Knowledge Base
AI agents in the same workspace can't learn from each other. We built a shared knowledge base using Obsidian-compatible vaults and MCP — here's why and how.
Your Claude Code Quota Is Burning 20x Faster Than You Think
A community investigation intercepted 17,610 API calls and found that invisible thinking tokens are the dominant quota cost in Claude Code. Six confirmed bugs, two still unfixed.
Real Estate Agents: Reclaim 2 Hours a Day from Admin Work
Listings, follow-ups, market research — the assistant handles it. You do the viewings.
How a Real Estate Agent Reclaimed 2 Hours a Day from Paperwork
Real estate agents lose 2+ hours daily to admin tasks that never close deals. Here's how an AI assistant handles the repetitive work — mandate tracking, visit reports, follow-ups — so agents can focus on clients.
How to Run an AI Agent in an Isolated VM
Sandboxes limit what your agent can do. A full Linux desktop in an isolated VM gives it everything — GUI, file system, persistence — without risking your host.
Automating Real Estate Listing Copy with an AI Assistant
Write property listings in 2 minutes instead of 30. Your assistant drafts the text, you review and publish.
Why every AI agent needs its own desktop
AI is evolving from chatbot to specialized agent with its own workstation. Here's why agent teams are the future, and how Le Bureau hosts them today.
Your AI agent reinstalls packages every session
Ephemeral sandboxes made sense for code execution. They fall apart for agents that do real work across sessions. Here's the persistence argument.
Sandboxes were step one. Agents need desktops now.
AI sandboxes solved code execution. But agents that browse, write documents, and work across sessions need something sandboxes can't provide: a real computer.
E2B vs Browserbase vs Scrapybara vs Le Bureau
A practical comparison of four AI agent platforms. What each one does, where it falls short, and how to pick the right infrastructure for your agent.
E2B alternatives for when your agent needs a real desktop
E2B handles code execution well, but agents that need browsers, persistent files, or GUI access need something different. A practical comparison of what's out there.
Why your ChatGPT blog posts don't rank
Google doesn't care that AI wrote your content. It cares that you skipped the 11 steps between 'generate text' and 'publish something worth reading.' Most AI content workflows are just copy-paste with extra steps.
Deploy your first AI agent desktop in 5 minutes
A quick walkthrough: create an account, spin up a cloud desktop, and give your AI agent its first task. BYOK, no local install.
Le Bureau vs E2B vs Scrapybara compared
A developer's comparison of three AI agent platforms with very different architectures and trade-offs.
Cloud desktops for AI agents: a practical guide
Why AI agents need full virtual machines, not sandboxes. Architecture, tradeoffs, setup, and what we learned building Le Bureau.
Monthly newsletter
Once a month. Not one more.
The best articles, product news, and one reading pick. Five minutes, start of the month.