CONTACTCMD

The AI-native Personal Relationship Manager for the command line.

Not a CRM. A PRM. Because the people in your network aren't leads in a pipeline. They're humans who trust you.

contactcmd
$ contactcmd gateway --review
Pending Approval
To Sarah Chen <sarah@acme.co> Via Email · Gmail API Agent follow-up-bot Subject Great meeting you at Demo Day Hey Sarah, Really enjoyed our conversation about the developer tools space yesterday. Would love to grab coffee next week and continue the chat — I had some thoughts on the API design we discussed. — Jason [a] approve [e] edit [d] deny
> a
✓ Approved · sent via Gmail API

The AI outreach industrial complex is here.

Every SaaS platform, every inbox tool, every "AI assistant" is racing to auto-send messages wearing your name. Mass personalization at scale. A/B tested empathy. Drip campaigns that pretend to care. They call it smart outreach. We call it what it is: sophisticated spam with your face on it.

And here's the thing nobody in the AI hype cycle wants to say out loud: your contacts didn't consent to this. Your mentor. Your college roommate. The investor who took a chance on your cold email three years ago. They built trust with you — not with your autonomous agent swarm.

ContactCMD exists because we believe the nuanced, personal decision tree — the one in your head that knows when to reach out, what to say, and when to just shut up — is what separates relationship building from spam. AI should serve that instinct, not replace it.

From CRM to PRM.

Customer Relationship Management was built for sales pipelines. Funnels. Conversion rates. It treats people like objects moving through stages. That made sense when the bottleneck was tracking. It doesn't make sense when the bottleneck is trust.

A Personal Relationship Manager flips that model. It's built around you — your voice, your judgment, your relationships. It uses relationship intelligence to remember context, surface the right moment, and draft with your tone. But the decision to reach out? That's always yours.

This is the difference between consent-first AI and autonomous spam. Between an AI firewall that protects your reputation and a bot that trades it for engagement metrics. Between treating your network like a database and treating it like what it actually is — people who chose to know you.

In an AI-first world, the human touch isn't a limitation. It's the whole point.

A zero-trust AI firewall between agents and your relationships.

The Gateway is your personal AI oversight layer — a localhost server that intercepts every AI-proposed message before it reaches anyone. Agentic workflows can draft emails, texts, and iMessages. But every message hits six layers of guardrails before it touches a real human. Nothing sends autonomously. Ever.

API Keys
HMAC-SHA256 per-agent authentication. Named keys. Full revocation. Zero-trust by default.
Rate Limits
Hourly and daily caps per key. No autonomous agent floods your contacts.
Allowlists
Per-key recipient scoping. Granular access control over exactly who each agent can reach.
Consent Flags
Per-contact AI opt-in. Consent-first by design. Your contacts choose whether they participate in agentic communication.
Content Filters
Regex and literal matching. Automated PII protection. Sensitive data never leaves the terminal.
You
Human-in-the-loop review. Edit, approve, or deny every message. The final guardrail is always a human decision.
01
Human-in-the-loop. No exceptions.
AI proposes. You approve, edit, or reject. Every single message. No autonomous sends. No "just this once." Your decision tree drives every interaction.
02
Local-first data sovereignty
SQLite on your machine. No cloud sync. No telemetry. No data broker. Your contact graph never leaves your hardware. Privacy by design, not by policy.
03
Authentic identity
Gmail API. iMessage. SMS. Always sent from your real accounts, in your real voice. Self-hosted and self-owned. Your reputation stays in your hands.
04
Relationship intelligence, not automation
Context-aware AI that surfaces the right moment — not a cron job that blasts your rolodex. Model-agnostic architecture. Works with any LLM. Serves your instincts instead of overriding them.

What's coming.

ContactCMD is the foundation. Here's everything we're building on top of it — every feature designed to make your relationships stronger without ever taking the wheel from you.

Privacy & Security
End-to-end encrypted local storage Coming Soon
Air-gapped mode Coming Soon
Privacy-compliant data export (GDPR) Coming Soon
Zero-telemetry architecture Coming Soon
Relationship Intelligence
Relationship graph visualization Coming Soon
Contact intelligence & enrichment Coming Soon
Interaction timeline & touchpoint tracking Coming Soon
Network health scoring Coming Soon
Smart cadence & timing optimization Coming Soon
AI & Agentic Workflows
On-device voice & tone calibration Coming Soon
Agentic workflow orchestration Coming Soon
Model-agnostic LLM support Coming Soon
AI transparency & explainability logs Coming Soon
Sentiment analysis guardrails Coming Soon
Integrations & Extensibility
CalDAV / CardDAV sync Coming Soon
Webhook & event API Coming Soon
Plugin & extension architecture Coming Soon
Multi-channel consent management Coming Soon
Self-hosted deployment options Coming Soon

One command. Your terminal.

$ brew install 5dayapp/tap/contactcmd
macOS · Rust-powered · Runs in your terminal