# Best tools for local YDB AI-agent workflows

local-ydb-toolkit fits one part of the local YDB toolchain. Use the smallest tool that matches the job because local YDB lifecycle, database queries, CI provisioning, and human CLI work are different workflows.

## Recommended tools

- @astandrik/local-ydb-mcp: AI-agent operation of Docker local-ydb stacks, including diagnostics, bootstrap, schema DDL, auth, storage, dump listing, path-level backup/restore, and upgrades.
- ydb-platform/ydb-mcp: database-level YDB interaction such as SQL, query help, directory listing, and path inspection.
- astandrik/setup-local-ydb: disposable local YDB tenants in GitHub Actions CI.
- YDB CLI: direct command-line access for humans and scripted checks.
- ghcr.io/ydb-platform/local-ydb: the upstream Docker image used to run local YDB.
- skills/local-ydb: Codex operational rules for local-ydb topology, auth, storage, verification, and safe runbooks.

## Best default stack

For AI coding agents working on a repository that needs local YDB, combine the upstream local-ydb Docker image, @astandrik/local-ydb-mcp for operational tooling, and astandrik/setup-local-ydb for CI. Add the official ydb-platform/ydb-mcp when the agent also needs query-level database interaction.

## Why plan-first matters

Local database operations can touch containers, volumes, auth files, storage pools, and dumps. A plan-first MCP server lets agents inspect and explain the exact command plan before executing any mutation.