📣 MarketClaw

AI marketing team orchestration plugin for OpenClaw.

MarketClaw turns a Telegram group into an autonomous AI marketing pipeline: Strategist → Creator → Reviewer → Publisher → Analyst — end to end.

---

Overview

MarketClaw forks DevClaw's battle-tested orchestration core and replaces the development pipeline with a content marketing pipeline. The same atomic state machine, heartbeat service, and audit trail — optimized for campaigns instead of code.

Campaign idea → Research brief → Content draft → Brand review → Publish → Analytics
     📊              🎯               ✍️               👁️           📣         📈
  Strategist       Strategist       Creator          Reviewer    Publisher   Analyst

---

Roles

Role	Levels	What they do
Strategist	junior / senior	CMO-level research. Writes campaign briefs, OKRs, creates content tasks
Creator	junior / medior / senior	Content writer. Posts, emails, articles. Opens content PR
Reviewer	junior / senior	Brand + quality review. Approves or rejects with specific feedback
Publisher	junior / senior	Publishes to platforms. Records published URLs and screenshots
Analyst	junior / senior	Collects metrics ~7 days post-publish. Feeds learnings into next brief

---

Installation

# Install MarketClaw
npm install -g @minglesai/marketclaw

# Add to openclaw.json plugins:
{
  "plugins": {
    "entries": {
      "marketclaw": {
        "module": "@minglesai/marketclaw"
      }
    }
  }
}

---

Quick Start

1. Set up MarketClaw (in Telegram group):

   /onboard
   

   Follow the guided setup. Creates workspace files and configures models.

2. Set up a campaign repo:

   git init my-campaigns
   cd my-campaigns
   mkdir -p campaigns content social/linkedin email reports
   cp /path/to/.agents/product-marketing-context.md .agents/

3. Register the project:

   /project_register name=mycampaigns repo=~/git/my-campaigns baseBranch=main
   

4. Fill in brand context (critical): Edit .agents/product-marketing-context.md with your company's voice, ICP, and positioning.

5. Launch a campaign:

   Create a campaign: "LinkedIn presence for our Q1 launch"
   

   MarketClaw creates an issue and dispatches a Strategist.

---

Repo Structure

After setup, your campaign repo looks like this:

my-campaigns/
├── .agents/
│   └── product-marketing-context.md  # 🚨 Fill this in — brand voice, ICP, positioning
├── campaigns/
│   └── <slug>/
│       ├── brief.md     # Strategist writes this
│       └── okrs.md      # Campaign OKRs
├── content/
│   └── <campaign>/      # Blog posts, articles
├── social/
│   ├── linkedin/
│   ├── twitter/
│   └── telegram/
├── email/               # Email campaigns
├── assets/              # Images, videos
└── reports/
    └── <campaign>/
        ├── urls.md       # Published URLs (Publisher writes this)
        └── metrics-*.md  # Analytics reports (Analyst writes this)

---

Workflow Pipeline

Planning ─── APPROVE ──→ To Do
                            │
                         PICKUP
                            ↓
                         Creating (Creator writes content + opens PR)
                            │
                         COMPLETE
                            ↓
                         To Review ←── auto-detected PR
                            │
                     APPROVED/SKIP
                            ↓
                         To Publish (PR merged, content in main branch)
                            │
                          PICKUP (Publisher worker)
                            ↓
                         Publishing (Publisher posts to platform)
                            │
                           PASS
                            ↓
                         Published (hold — waiting 7 days)
                            │
                    [heartbeat cron after 7 days]
                            ↓
                         To Analyze → Analyzing → Done

Revision loop: Any step can route to To Improve → Creating for fixes.

---

Configuration

Edit marketclaw/workflow.yaml in your workspace:

workflow:
  reviewPolicy: human    # human | agent | skip
  publishPolicy: agent   # agent | skip | human
  analyzeAfterDays: 7    # Days after publish before analytics
  
roles:
  creator:
    models:
      medior: anthropic/claude-sonnet-4-5
  publisher:
    models:
      junior: anthropic/claude-sonnet-4-5

---

Content File Format

All content files use frontmatter:

---
platform: linkedin
campaign: q1-launch
publish_date: "2024-03-15"
hashtags: ["#AI", "#SaaS"]
status: draft
---

Content goes here...

See defaults/marketclaw/CONTENT_SPEC.md for the full spec.

---

Tools

MarketClaw registers the same tools as DevClaw with marketing-appropriate descriptions:

Tool	What it does
task_create	Create campaign issues
task_start	Advance issue to next queue
research_task	Dispatch Strategist for campaign research
work_finish	Complete a worker's task
tasks_status	Live dashboard of all campaign work
project_register	Register campaign repo
setup	Configure MarketClaw

---

License

MIT — MinglesAI