Zero to CRM for Developers: Build Your Own in an Afternoon

Developers don't need to use off-the-shelf CRMs. With DenchClaw, you can build a fully custom CRM tailored to your workflow in a single afternoon.

Mark Rachapoom

March 26, 2026·7 min read

Zero to CRM for Developers: Build Your Own in an Afternoon

Most developers I know use a CRM badly or not at all. Either they're paying for HubSpot and ignoring half the features, or they're tracking everything in a Google Sheet that never gets updated. Neither works.

The problem isn't motivation. It's that off-the-shelf CRMs aren't designed for how developers think. They're designed for sales managers. They have too many features in the wrong places, not enough SQL access, and zero extensibility.

DenchClaw is different. It's a CRM built for people who'd rather write a query than click a button. And crucially: you can have a fully working, custom CRM running in an afternoon.

Here's how.

Step 1: Install DenchClaw (5 minutes)#

npx denchclaw

This installs the OpenClaw agent runtime with the dench profile, starts a gateway on port 19001, initializes a DuckDB workspace, and opens localhost:3100 in your browser.

The first run will ask you a few setup questions. Answer them, or skip them — you can configure everything through the agent later.

Step 2: Define Your Objects (15 minutes)#

An Object is what other CRMs call a table. Tell the agent what you want to track:

"Create a CRM with three objects: contacts, companies, and deals. Contacts should have: name, email, phone, company (relation), role, LinkedIn URL, last contacted date, and status (Lead, Active, Churned). Companies should have: name, website, industry, employee count, and ARR. Deals should have: title, value, stage (Prospect, Qualified, Proposal, Closed Won, Closed Lost), contact (relation), company (relation), close date, and notes."

The agent will:

Create the objects in DuckDB
Write the .object.yaml files for each
Generate PIVOT views (v_contacts, v_companies, v_deals)
Add the objects to your sidebar navigation

In under 15 minutes you have a fully functional CRM schema. No migrations to run. No admin panel to navigate.

Step 3: Set Up Your Views (15 minutes)#

By default, each object shows a Table view. Tell the agent how you want to see your data:

"Set up a Kanban view for deals, grouped by Stage. Create a filtered table view for contacts showing only Leads, sorted by Last Contacted ascending — call it 'Stale Leads'. Create a calendar view for deals based on Close Date."

Each view is saved as a configuration in the .object.yaml. The frontend updates immediately.

You can also set up views manually by editing the YAML:

views:
  - name: Stale Leads
    type: table
    filters:
      - field: Status
        operator: equals
        value: Lead
    sort:
      - field: Last Contacted
        direction: asc
  - name: Pipeline
    type: kanban
    kanbanField: Stage

Step 4: Import Your Existing Data (20 minutes)#

You probably have data somewhere already — a spreadsheet, a CSV export, a Notion database. Import it:

# Via the agent
"Import this CSV file as contacts: ~/Desktop/contacts.csv"
 
# Or directly via DuckDB
duckdb ~/.openclaw-dench/workspace/workspace.duckdb << 'EOF'
-- First, check what you're working with
SELECT * FROM read_csv_auto('/Users/yourname/Desktop/contacts.csv') LIMIT 5;
 
-- Then insert into the contacts object
INSERT INTO entries (object_id, created_at)
SELECT
  (SELECT id FROM objects WHERE name = 'contacts'),
  current_timestamp
FROM read_csv_auto('/Users/yourname/Desktop/contacts.csv');
EOF

The agent can also import from HubSpot, Notion, or any site you're logged into, by opening it in the browser and exporting the data automatically.

Step 5: Connect Your Messaging Channel (10 minutes)#

DenchClaw's power comes from being able to talk to your CRM from anywhere. Set up Telegram:

Message @BotFather on Telegram
Create a bot: /newbot → give it a name → copy the token
In DenchClaw: "Connect my Telegram bot with token [your-token]"

Now you can query your CRM from your phone:

"How many open deals do I have?" "Add a new contact: Sarah Chen, Head of Product at Stripe, sarah@stripe.com" "Show me my stale leads" "Log a meeting with John from Acme today"

Step 6: Build a Custom Dashboard (20 minutes)#

This is where it gets interesting. DenchClaw has an app builder — you can build custom web apps with full DuckDB access that appear in your sidebar.

Ask the agent:

"Build me a pipeline dashboard that shows: (1) count of deals by stage as a bar chart, (2) total deal value by stage, (3) deals closing this month, and (4) my top 5 contacts by deal value. Use Chart.js."

The agent will build a .dench.app with:

<!-- apps/pipeline-dashboard.dench.app/index.html -->
<canvas id="stageChart"></canvas>
<script>
const stageData = await dench.db.query(`
  SELECT "Stage", COUNT(*) as count, SUM("Value") as total
  FROM v_deals
  WHERE "Stage" NOT IN ('Closed Lost')
  GROUP BY "Stage"
  ORDER BY total DESC
`);
// ... Chart.js rendering
</script>

The app appears in your sidebar immediately, auto-refreshes when data changes.

Step 7: Add Action Buttons (15 minutes)#

Action fields are buttons on each row that run server-side scripts. This turns your CRM from a database viewer into an action launchpad.

Ask the agent:

"Add an 'Enrich' button to contacts that, when clicked, looks up their email on Apollo.io and fills in their company, role, and LinkedIn URL"

Or add an email button:

"Add a 'Send Follow-up' action to the Stale Leads view that drafts a personalized follow-up email for each contact using their deal history and last contact date"

The action script runs on click, streams results back to the UI, and updates the entry.