Dex Skill for AI Agents
The Dex Skill is an installable package that gives AI agents (like Claude Code) the knowledge and tools to manage your personal CRM. It combines tool access with CRM best practices — so your AI assistant doesn’t just execute commands, it understands how to be an effective relationship manager.MCP Server vs. Skill: The MCP Server provides the raw tools. The Skill wraps those tools with CRM expertise — meeting prep workflows, follow-up cadences, organizational strategies, and more. For the best experience, use both together.
Installation
- Claude Code
- Skills CLI
- Other AI Agents
Setup
On first use, the skill runs a setup script that automatically picks the best connection method for your environment:MCP tools detected
If
dex_* tools are already available (e.g., via the MCP Server), the skill uses them directly. No additional setup needed.Neither available — run setup
The skill runs
bash scripts/setup.sh which auto-detects your environment and picks the best path:| Environment | Method | Requires |
|---|---|---|
| Desktop with Go installed | Generates CLI binary via CLIHub + OAuth | Go 1.26+, browser |
| Desktop with Node.js | Configures hosted MCP server via add-mcp | Node.js, browser |
| SSH / headless with TTY | Device code flow (enter code on any device) | Internet access |
| CI / automation | Pre-built binary + API key | DEX_API_KEY env var |
Headless & CI Setup
For servers and CI pipelines where no browser is available:- Device Code Flow (Interactive)
- API Key (Non-Interactive)
- OpenClaw
If you have terminal access (SSH), the setup script automatically starts a device code flow:
- Downloads the CLI binary for your platform
- Displays a short code (e.g.,
ABCD-1234) and a URL - Open the URL on any device, log into Dex, and enter the code
- The CLI receives an API key and completes setup
Platform Support
Pre-built CLI binaries are available for:| OS | Architecture |
|---|---|
| macOS | Apple Silicon (arm64), Intel (amd64) |
| Linux | amd64, arm64 |
| Windows | amd64, arm64 |
Core Workflows
The skill teaches your AI assistant these CRM workflows:1. Find a Contact
Ask your agent to look up anyone by name, email, company, or any keyword. “Look up John from Acme Corp” “Who do I know at Google?” “Find my contacts tagged as Investor” The agent searches your contacts and retrieves full details including interaction history when needed.2. Add a New Contact
“Add Jane Doe — she’s VP Engineering at Acme Corp, email jane@acme.com” The agent creates the contact and can immediately organize them with tags, groups, and reminders.3. Log an Interaction
“Log that I had a coffee meeting with Jake today. We discussed the Series A and he’s targeting Q3.” The agent creates a timestamped note on the contact’s timeline with the appropriate note type (Meeting, Call, Coffee, etc.).4. Set a Reminder
“Remind me to follow up with Maria in two weeks about the proposal” The agent creates a reminder linked to the contact with the right due date. Supports recurring reminders:weekly, biweekly, monthly, quarterly, biannually, yearly.
5. Prep for a Meeting
“I have a meeting with Sarah Chen tomorrow — prep me” The agent generates a meeting brief with: - Who they are (role, company, relationship context) - Last interaction (when, what was discussed) - Pending items (open reminders, action items) - Suggested talking points based on your history6. Organize Your Network
“Tag everyone I met at the TechCrunch conference as ‘TC Disrupt 2026’” “Create a group called ‘Advisory Board’ and add Marcus and Lisa” Tags are flat labels for cross-cutting categories (e.g., “Investor”, “College Friend”). Groups are named collections with emoji and description (e.g., ”🚀 Startup Advisors”).7. Track Custom Data
“Create a custom field called ‘Deal Stage’ with options: Prospect, Qualified, Negotiation, Closed” “Set Jake’s deal stage to Qualified” Three field types: free text (input), dropdown (autocomplete), and date (datepicker).