Your CRM as a Second Brain: Beyond Contacts and Deals

I've been thinking about the word "CRM" and how limiting it is.

Customer Relationship Management. The name implies a narrow scope: customers, and your relationships with them. Contacts table. Deals pipeline. Activity log. That's the mental model most people carry.

But the underlying capability — a structured database you can query in natural language, backed by an AI agent that remembers everything — is vastly more powerful than the name suggests. DenchClaw is what happens when you take that capability and refuse to limit it to just "customers."

What "Second Brain" Means Here#

The second brain concept (popularized by Tiago Forte) is about externalizing your mental context: the things you're tracking, the decisions you've made, the knowledge you've accumulated. The goal is to free your actual brain from the overhead of remembering, so it can focus on thinking.

Most second brain implementations are document-centric: Notion, Obsidian, Roam Research. Write notes, link them, build a knowledge graph. This works, but it has limits: notes require curation, don't structure well for querying, and don't talk back.

DenchClaw as a second brain is different: it's structured, queryable, and backed by an AI agent that has full context on everything you've stored. The combination is more powerful than either a document store or a traditional CRM.

What You Can Track Beyond Contacts and Deals#

DenchClaw lets you create any Object — any table you want. This means your second brain can include:

Projects and Tasks#

Create a projects object with fields: Name, Status (Active, On Hold, Completed), Priority, Owner, Start Date, End Date, Description. Create a tasks object with fields: Title, Status, Project (relation to projects), Due Date, Assignee.

Now ask: "What projects are stuck? Which tasks are overdue? What's my completion rate this quarter?"

Investments and Portfolio Tracking#

Create an investments object: Company, Stage, Check Size, Date, Lead Investor, Status (Active, Exited, Written Off), Notes. Link to a founders object.

Ask: "What's my total deployed capital? Which companies haven't given me an update in 90 days? Which are pre-Series A?"

Hiring Pipeline#

Create a candidates object: Name, Role, Stage (Applied, Phone Screen, Interview, Offer, Hired, Rejected), Resume URL, Notes, Interviewer, Last Update.

Ask: "How many candidates are in interview stage for the engineer role? What's my offer conversion rate this quarter?"

Research Notes#

Create a research object: Topic, Source, Date, Key Insight, Tags, Related Contacts. Link to people when research is about specific individuals or companies.

Ask: "What have I learned about distributed systems in the last 3 months? What research is relevant to my meeting with Databricks tomorrow?"

Personal Network (beyond customers)#

Create an advisors, investors, mentors object — distinct from customers but tracked with the same rigor. Track last contact, context from last conversation, what you want to discuss next.

Product Decisions#

Create a decisions object: Title, Date Made, Decision, Rationale, Alternatives Considered, Outcome. This is institutional memory — why you made the choices you did.

Ask: "Why did we decide to use EAV over a traditional schema? What were the alternatives?"

The AI Layer Makes This Powerful#

Any of these as standalone databases would be useful. What makes DenchClaw genuinely a second brain is the AI agent layer.

Cross-object context: When you're preparing for a meeting with an investor who is also tracked as an advisor in your portfolio, the agent has context from all three Objects — investor record, advisor notes, investment tracking — without you having to manually surface it.

Natural language queries: "Show me all the people I've talked to at companies in the developer tools space, sorted by when I last spoke to them" is a query that spans your contacts, companies, and activity data. Write the SQL if you want, or just ask in English.

Proactive reminders: The agent's heartbeat checks can surface anything: "You have 3 investments that haven't filed updates in 90 days." "Your hiring pipeline has 5 candidates in 'Phone Screen' stage for more than 2 weeks." "You haven't updated your research notes on LLM architecture in 6 months."

Memory across sessions: The agent's MEMORY.md and daily logs capture important context. When you tell the agent "remember that I'm planning to raise in Q4 at a $15M pre-money valuation," that context persists and can inform future answers.

Entry documents: Every entry can have a rich markdown document. Your meeting notes, your research write-ups, your project retrospectives — all linked to the relevant structured record and accessible by the agent.

The Knowledge Architecture#

Here's how I think about organizing a DenchClaw second brain:

People layer: Everyone you know or have known — customers, investors, advisors, colleagues, founders, researchers. One unified people object with status and tags to differentiate.

Organizations layer: Companies, universities, research labs, VCs, communities. Linked to people.

Things layer: Projects, deals, investments, properties, products — whatever structured "things" you're tracking. Linked to relevant people and organizations.

Knowledge layer: Documents, research notes, decisions, meeting notes. These live primarily as entry documents and markdown files in the workspace, searchable by the agent.

Time layer: Events, activities, follow-ups. The agent maintains these implicitly through the heartbeat system and can surface them explicitly.

Getting Started: Building Your Second Brain#

The practical path:

1. Start with what you're already tracking: If you have a spreadsheet of contacts, import it. If you have a Notion database, ask the agent to import it.

2. Add one new Object at a time: Don't try to model everything up front. Add "projects" when you feel the pain of not having it tracked.

3. Let the agent help with structure: "I want to track my investment portfolio. What fields should I include?" The agent will suggest a sensible schema.

4. Write entry documents: When you have a significant meeting, ask the agent to write up the notes and link them to the relevant contact and company. This is where the knowledge accumulates.

5. Review regularly: Ask the agent "what should I be paying attention to this week?" and let it synthesize across your entire second brain.

Frequently Asked Questions#

Is DenchClaw a replacement for Notion or Obsidian?#

It's complementary. Notion and Obsidian are better for document-heavy workflows, long-form writing, and public wikis. DenchClaw is better for structured data with AI querying and action capabilities. Many people use both: Notion for team docs, DenchClaw for personal relationship and business data.

How does the agent know what to remember?#

The agent reads its MEMORY.md and recent daily log files at the start of each session. It updates MEMORY.md over time with important context you've shared. You can also explicitly tell it: "Remember that [X]."

Can I link documents from Obsidian or Notion into DenchClaw?#

Not directly, but the agent can import content from both (Obsidian via file access, Notion via browser). You can also link external URLs in entry documents.

How private is the second brain?#

Completely private. All data is on your machine. The AI model API calls are the only data that leaves your machine, and those are the content of questions you ask — not your underlying CRM data.

Ready to try DenchClaw? Install in one command: npx denchclaw. Full setup guide →