Ana Sofia Grass Giraldo shows how to connect Claude Code to Notion via MCP and build a full operating system for your company, personal work, and GTM engine, all through conversation.
Guest Speaker
Product Owner, GTM Engineering & BI at Kapwork
Ana Sofia Grass Giraldo is a Systems Engineering and Data Science student at Pontificia Universidad Javeriana in Colombia, where she earned a full merit scholarship for the highest GPA in her Data Science program. She now works as Product Owner of GTM Engineering and Business Intelligence at Kapwork, where she built the company's entire GTM stack from scratch using Claude, Clay, and Notion. She also trains professionals in AI-powered GTM automation at Bootprint. Before any of this, she wrote her thesis with CERN's ATLAS experiment, building a mobile tool that registered 1,600 temperature sensors into production, a project that earned a formal letter of recognition from the ATLAS ITk coordinator. She also founded CiberAbu, a startup teaching digital literacy to senior citizens in Colombia.
Session Overview
Context lives inside people's heads, scattered across chats, docs, and spreadsheets. Scaling adds more people, which adds more mess.
Personal command center, company operating system, and GTM engineer sandbox. All built through conversation with Claude Code.
Notion for shared operations, Obsidian for your personal Claude brain, GitHub for code and versioning. Each tool has a specific job.
Trend spotting agents, content calendars, live CRM pipelines, and scheduled weekly reports built with Claude routines.
Live Demo
Install Claude Code, tell it to help you connect the Notion MCP, and Claude walks you through the steps. One connection gives you full access to create, edit, translate, and organize everything in your Notion workspace through conversation.
A system for projects, memory, tasks, and content in one Notion space. Claude built the entire structure from a prompt, not a template.
An executive roadmap covering mission, vision, quarterly planning, and brand identity. A people and roles database where each person links to their tasks. An execution hub for tasks, meetings, clients, and products. A mission control dashboard where every metric calculates itself automatically.
A trend spotting agent connected via Claude routines that pulls from newsletters daily and drafts LinkedIn posts (one reached 25,000 views in four days). A content calendar with a personal brand voice baked in. A live CRM pipeline that updates from real data instead of manual entry. Scheduled weekly reports that run by themselves when Sofia is offline.
Obsidian stores all of Sofia's Claude context: documents, guides, research, custom rules. Because it syncs to local files, Claude can reference it without re-reading everything fresh each session. That's what keeps token usage manageable even on a $200/month Max plan with five-plus hours of daily use.
Mental Models
Three tools with three distinct jobs. Notion is cloud-based and shareable so the whole team can work in it. Obsidian holds your personal Claude context so you don't repeat yourself every session. GitHub handles code, versioning, and anything the engineering team needs to collaborate on.
Context first. When Claude already knows your company's mission, voice, people, and data, every prompt you write gets shorter and every output gets sharper. The OS is not setup overhead. It's the foundation that everything else runs on.
Every time you re-explain context from scratch, you're burning tokens and slowing yourself down. Obsidian is your persistent memory layer. Store your context once, reference it every session. On a lower plan, skills from the Claude Code skills directory can help you stretch further without re-pasting the same prompts.
Obsidian vaults are local. You can't connect your Obsidian vault to your boss's Obsidian vault. Notion is shareable, has relational databases, lets you build multiple views of the same data (calendar, gallery, chart), and gives the whole team live access. If more than one person needs it, Notion wins.
Example Prompts
From the Session
Why go through Claude instead of just using Notion AI?
Are you using Notion agents, or external agents connected through Claude Code?
Does running all this context through MCP burn through your Claude usage limits fast?
If you could only keep one integration between Claude Code and Notion, which one would you keep?
Why still use Obsidian if Notion already does so much?
How do you avoid burning too many tokens when Claude has to dig through all that context?
Tools
Resources
Full session on Zoom, approximately 60 minutes
The presentation Sofia walked through, built and deployed with Claude Code and Vercel
One week. Hands-on. Small group. You leave with a working Claude Code system and the skills to keep building on your own.
Join the Claude Code CohortStay in touch