SQLite — The Power Move
Native database read/write. Dense structured data. SQL queries via natural language.
Markdown files are great for narrative context — project descriptions, decision logs, preference lists. But some information is inherently structured: customer records, API endpoints, product catalogs, configuration tables, metric histories. For this kind of data, SQLite is the power move.
OpenClaw has native SQLite support. Your agent can create databases, define tables, insert records, and run queries — all through natural language. Say "create a customer database with columns for name, email, company, last contact date, and status" and your agent creates the SQLite file, defines the schema, and confirms it's ready. Say "add John Smith, john@example.com, Acme Corp, contacted last Tuesday, status active" and the record is inserted.
The real power is in queries. "Show me all customers I haven't contacted in 30 days." "What's the most common status among customers from companies with more than 5 contacts?" "Give me a summary of all API endpoints grouped by service." Your agent translates natural language into SQL, runs the query, and presents the results. No SQL knowledge required on your part.
Why SQLite instead of a "real" database? Locality. The SQLite file lives in your workspace alongside your memory files. No server to manage, no connection strings, no authentication. It's a single file. It travels with your workspace. It's backed up with your workspace. And for the scale most individual users need — thousands to hundreds of thousands of records — SQLite is more than sufficient.
The pattern that emerges: use structured folders for narrative context (descriptions, logs, preferences), Mem0 for automatic conversational capture (decisions, quick facts), and SQLite for dense structured data (records, tables, metrics). Each tool excels at a different type of information, and together they give your agent comprehensive memory across all data types.
Walkthrough
- 1Verify SQLite tools are enabled in your
openclaw.json - 2Ask your agent to create a test database: 'Create a SQLite database called contacts.db with a table for name, email, company, and notes'
- 3Insert 5 test records using natural language
- 4Run queries: 'Show all contacts', 'Find contacts from [company name]', 'Count contacts by company'
- 5Test updates: 'Update John Smith's notes to include that we discussed pricing on March 15'
- 6Verify the database file exists in your workspace:
ls *.db - 7Practice a real use case: create a database for something relevant to your work (projects, leads, content ideas, etc.)
- 8Run a complex query that demonstrates the value over flat files
Checkpoint
You've created a SQLite database, inserted records, run natural language queries, and identified a real use case where structured database storage is better than markdown files.