crm.cli

A headless, CLI-first CRM for AI native companies. Contacts, deals, and pipeline in a single SQLite file — queryable from your terminal, composable with Unix tools, and mountable as a virtual filesystem so any tool that reads files (Claude Code, Codex, grep, jq, vim) has full CRM access without any integration.

No server. No Docker. No accounts. No GUI. Just npm install -g @dzhng/crm.cli and go.

Created by Duet — a cloud agent workspace with persistent AI. Set up crm.cli in your own private cloud computer and run it with Claude Code or Codex — no local setup required. Try Duet →

Why crm.cli

Existing CRMs are GUI-first tools built for sales teams. If you're a technical founder, indie hacker, or engineer running BD, you're probably managing contacts in a spreadsheet you grep through. crm.cli is built for you.

Your CRM is a filesystem. Mount it with crm mount ~/crm and every tool that reads files — AI agents, shell scripts, editors — gets full CRM access for free. No MCP servers, no API keys, no integration code. The filesystem is the universal API.

crm mount ~/crm
ls ~/crm/contacts/
cat ~/crm/contacts/jane-doe.json | jq .name
# Point Claude Code at ~/crm and ask it to research your pipeline

Deep data normalization. A CSV has zero setup cost. crm.cli justifies its existence with structured intelligence: E.164 phone normalization (look up by any format), website normalization, social handle extraction (paste a LinkedIn URL, it stores the handle), entity merge with reference relinking, and fuzzy duplicate detection. This is what a spreadsheet can never provide.

crm contact add --name "Jane Doe" \
  --email jane@acme.com \
  --phone "+1-212-555-1234" \
  --linkedin linkedin.com/in/janedoe \
  --company Acme
# Phone stored as E.164, LinkedIn URL → handle, company auto-linked

Pipe-friendly by default. Every command outputs structured data. Pipe to jq, grep, awk, or feed into scripts. --format json on everything.

crm deal list --stage qualified --format json | jq '.[] | .value'
crm find "that fintech CTO from London"
crm report pipeline
crm dupes --threshold 0.5

Install

npm install -g @dzhng/crm.cli
# or: bun install -g @dzhng/crm.cli

# Or install the compiled binary
curl -fsSL https://raw.githubusercontent.com/dzhng/crm.cli/main/install.sh | sh

Install as an AI agent skill

Point your agent at the skills folder to give it full CRM knowledge — install instructions, every command, the filesystem interface, and tips for agents:

# Claude Code / Duet — install the skill from GitHub
claude skills add https://github.com/dzhng/crm.cli/tree/main/skills

# Or copy skills/SKILL.md into your agent's skills directory
cp skills/SKILL.md ~/.your-agent/skills/crm-cli/SKILL.md

Storage

Everything lives in a single SQLite file. Default: ~/.crm/crm.db.

crm --db ./my-project.db contact list    # use a specific database
export CRM_DB=./team.db                  # or set via env var

No server. No Docker. No accounts. Back it up by copying the file.

Configuration

Config is loaded from crm.toml. Resolution order (first match wins):

1. --config flag (explicit)
2. CRM_CONFIG env var
3. Walk up from CWD: ./crm.toml → ../crm.toml → ../../crm.toml → ... → /crm.toml
4. ~/.crm/config.toml (global fallback)