home / skills / openclaw / skills / crm-in-a-box

crm-in-a-box skill

safe

This skill helps you bootstrap and manage an open file-based CRM, log contacts and interactions, and track deals without SaaS.

npx playbooks add skill openclaw/skills --skill crm-in-a-box

Review the files below or copy the command above to add this skill to your agents.

Files (5)

SKILL.md

2.7 KB

---
name: crm-in-a-box
description: Bootstrap and manage an open, file-based CRM using the CRM-in-a-Box protocol (contacts, pipeline, interactions as NDJSON). Use when setting up a new CRM for a company, logging contacts/leads, updating deal stages, or recording interactions — all without any SaaS dependency.
metadata:
  {
    "openclaw":
      {
        "emoji": "📇"
      }
  }
---

# CRM-in-a-Box Skill

An open, agent-native CRM protocol. One directory = one CRM. No vendor lock-in.

## Files

- `contacts.ndjson` — one JSON object per line, each a contact/company record
- `pipeline.ndjson` — deal/opportunity tracking with stages
- `interactions.ndjson` — append-only log of emails, calls, notes, meetings
- `config.yaml` — pipeline stages and labels

## Bootstrap a New CRM

Copy the baseline files into a company directory:

```bash
mkdir -p ./mycompany/crm
cp /path/to/crm-in-a-box/{config.yaml,contacts.ndjson,pipeline.ndjson,interactions.ndjson} ./mycompany/crm/
```

Or start fresh with empty NDJSON files and the default config.

## Contact Schema

```json
{
  "id": "c001",
  "name": "Jane Smith",
  "email": "[email protected]",
  "company": "Acme Corp",
  "phone": "+1-555-0100",
  "stage": "new",
  "labels": ["hot-lead"],
  "notes": "Referred by John.",
  "created_at": "2026-01-01T00:00:00Z"
}
```

## Pipeline Schema

```json
{
  "id": "p001",
  "contact_id": "c001",
  "stage": "proposal_sent",
  "deal": "Enterprise License",
  "value": 12000,
  "updated_at": "2026-01-15T00:00:00Z"
}
```

## Interaction Schema

```json
{
  "id": "i001",
  "contact_id": "c001",
  "type": "email",
  "summary": "Sent intro email about Denver market sale.",
  "at": "2026-01-15T10:00:00Z"
}
```

## Pipeline Stages (default)

`new` → `contacted` → `meeting_scheduled` → `proposal_sent` → `negotiating` → `won` / `lost`

## Default Labels

`hot-lead`, `warm-lead`, `cold-lead`, `referral`, `conference`, `inbound`, `outbound`

## Agent Instructions

- **Log a contact:** append a JSON line to `contacts.ndjson`
- **Update a stage:** append an updated entry to `pipeline.ndjson` (keep old entries — append-only)
- **Log an interaction:** append to `interactions.ndjson`
- **Search contacts:** `grep -i "name" contacts.ndjson | python3 -m json.tool`
- **List pipeline:** `cat pipeline.ndjson | python3 -c "import sys,json; [print(json.dumps(json.loads(l), indent=2)) for l in sys.stdin]"`

## Forks / Verticals

Fork `config.yaml` to customize stages and labels for your vertical:
- `pm-crm` — Property management (tenants, owners, vendors)
- `saas-crm` — SaaS sales
- `realestate-crm` — Buyers, sellers, listings
- `recruiting-crm` — Candidates, jobs, placements

---

*Part of the [biz-in-a-box](https://biz-in-a-box.org) family of open protocols.*

Overview

This skill bootstraps and manages an open, file-based CRM using the CRM-in-a-Box protocol. One directory equals one CRM: contacts.ndjson, pipeline.ndjson, interactions.ndjson and a config.yaml define the system. It avoids SaaS lock-in by recording everything as append-only NDJSON so data is portable and auditable. Use it to stand up a lightweight, agent-friendly CRM for teams that prefer plain files.

How this skill works

The skill creates or copies the baseline files into a company directory and enforces simple NDJSON schemas for contacts, pipeline entries, and interactions. All operations are append-only: add a contact by appending a JSON line to contacts.ndjson, update deal stages by appending a pipeline entry, and log activities to interactions.ndjson. A configurable config.yaml defines pipeline stages and labels so you can fork for verticals. Simple command-line patterns (grep, python json.tool) let agents and scripts search and render records.

When to use it

Setting up a new CRM for a small company or project without relying on third-party SaaS
Logging new contacts and leads in a reproducible, auditable format
Tracking deal progress with append-only pipeline entries to preserve history
Recording calls, emails, and meeting notes as a tamper-evident interaction log
Creating vertical- or team-specific CRMs by forking config.yaml

Best practices

Keep NDJSON files under version control or periodic backups to preserve history and enable restores
Treat all updates as append-only; never rewrite existing lines to maintain an audit trail
Use stable, unique IDs (e.g., c001, p001, i001) and ISO 8601 timestamps for consistency
Customize config.yaml per vertical to reflect real workflow stages and sensible labels
Use simple CLI filters and JSON pretty-printing for quick inspection; add small scripts for common queries

Example use cases

Bootstrap a company CRM by copying the baseline files into ./mycompany/crm and start appending leads
Log an inbound lead: append a contact with labels like inbound and hot-lead, then create an initial pipeline entry
Move a deal through stages by appending new pipeline entries (proposal_sent → negotiating → won) to preserve history
Record all customer interactions (emails, calls, meetings) in interactions.ndjson to keep a unified activity log
Fork config.yaml into saas-crm or realestate-crm to match domain-specific stages and labels

FAQ

Can I search records quickly from the command line?

Yes. Use grep and python3 -m json.tool or small Python scripts to filter and pretty-print NDJSON lines.

How do I change pipeline stages or labels?

Edit or fork config.yaml for your vertical. Existing entries remain in NDJSON; new entries should follow the updated stages and labels.