How to Use OpenClaw with MongoDB — Agent Data Persistence and

MongoDB is a natural fit for OpenClaw agents because it handles semi-structured data well — and agent-generated data is almost always semi-structured. Conversation summaries, scraped data, lead records, task results — all of it lands cleanly in documents rather than rigid SQL tables.

When to Use MongoDB with OpenClaw

• Storing structured outputs from research or scraping tasks
• Building a CRM or lead database your agent can query conversationally
• Persisting agent task history beyond workspace files
• Running analytics queries over agent-collected data

Step 1: Set Up MongoDB Atlas

1. Create account at mongodb.com/atlas
2. Create a free M0 cluster (512MB)
3. Get your connection string: mongodb+srv://user:pass@cluster.mongodb.net/agentdb
4. Add to ~/.openclaw/.env: MONGODB_URI=mongodb+srv://...

Step 2: Create a MongoDB Skill

Create ~/.openclaw/workspace/skills/mongodb/SKILL.md:

# MongoDB Skill
Use mongosh for all database operations.

## Insert Document
mongosh "$MONGODB_URI" --eval 'db.leads.insertOne({name:"$NAME",email:"$EMAIL",source:"openclaw",createdAt:new Date()})'

## Query Documents
mongosh "$MONGODB_URI" --eval 'db.leads.find({source:"openclaw"}).limit(10).toArray()'

## Update Document
mongosh "$MONGODB_URI" --eval 'db.leads.updateOne({email:"$EMAIL"},{$set:{status:"contacted"}})'

Step 3: Teach Your Agent When to Write

In your AGENTS.md:

## Data Persistence
When you collect structured data (leads, research, task outputs),
store in MongoDB. Always include:
- source: "openclaw"
- agentId: your agent name
- createdAt: current timestamp

Step 4: Conversational Queries

Ask your agent natural language questions like "how many leads did we get this week?" — it translates to a query, runs it, and answers in plain English. Configure this in SOUL.md:

When asked about data or counts, query MongoDB first rather than estimating.
Always show the query you ran alongside the answer.

Step 5: Index for Performance

db.leads.createIndex({createdAt: -1})
db.leads.createIndex({source: 1, status: 1})

Security Notes

• Use a dedicated MongoDB user with minimum required permissions
• Enable IP allowlisting in Atlas — only add your VPS/host IP
• Never log the full connection string — use env vars only

Want the complete MongoDB + OpenClaw pattern including multi-collection schemas and aggregation pipelines? The OpenClaw Playbook ($9.99) covers production-grade data architecture for AI agents.