Plan a safe data backfill
Confirm schema, scope, and a read-back assertion before any UPDATE; retain a verification artifact after the approved write.
Read the safe-backfill guide →Agent database playbooks
For every database task,
start with enough evidence.
These are not command encyclopedias. They show how dbcli helps an agent confirm structure, respect boundaries, and deliver reviewable results in real work.
Published guides
Start with the task, not a command
Every guide covers the risk, a repeatable workflow, the evidence it delivers, and the boundary dbcli will not cross for you.
Deliver a shareable data dashboard
Turn a business question into a bounded query and interactive HTML report without asking an agent to guess the columns.
Read the dashboard guide →Find ORM-to-database schema drift
Compare the live schema with ORM definitions before the team decides whether any DDL should run.
Read the schema-drift guide →Assess a schema change offline
Write a bounded report of known governed dependencies and visible coverage gaps before migration review.
Read the impact-assessment guide →Verify an approved write and retain provenance
Read back the expected state and record a safe receipt without treating evidence as authorization.
Read the verification-evidence guide →Prepare an evidence pack for review
Compose, validate, and render a compact integrity-protected handoff from existing evidence.
Read the evidence-pack guide →Protect governed business terms with contracts
Keep owners and descriptive evidence expectations visible beside semantic metadata.
Read the semantic-contracts guide →Use database evidence to find a slow API
Narrow the problem through observed events, query plans, and index guidance instead of tuning production on instinct.
Read the slow-endpoint guide →When should vibe coding use dbcli?
Use a visual guide to choose between a direct client, MCP, and dbcli—and know when a local experiment needs a repeatable workflow.
Read the tool-choice guide →