dbmd-cli
The dbmd command-line tool for db.md, the open standard for
databases in plain files.
db.md is a database made of markdown files: records are markdown with
YAML frontmatter, relationships are wiki-links, the directory is the
database, frontmatter carries structured fields, DB.md declares schemas,
and your agent is the query engine. For semantic, evolving, workflow-heavy
software, the old database/backend/frontend stack can become readable context,
an agent harness, and a generated surface. dbmd is a
single deterministic binary that performs every db.md file/data operation; all
logic lives in
dbmd-core and the binary is a
thin arg-parse/format wrapper.
Install
This installs the dbmd binary. Alternatives: brew install carloslfu/tap/dbmd, or a prebuilt, checksummed, provenance-attested
tarball from the GitHub releases.
Use
# create a store — you write DB.md (the agent authors it; there is no `dbmd init`)
# operate it
# load the canonical contract — the single source of truth, read once per session
To persist the contract across sessions, drop the open-format skill
skills/db-md/SKILL.md where your harness reads skills (copy it, use your
harness's own skill installer, or tell your agent) — there is no dbmd
install command; the installer is text.
Every subcommand supports --json and --help; none prompt
interactively.
Design
- Zero AI dependencies. No model calls, no embeddings, no vectors — ever. Your own agent harness (Claude Code, Codex, or any tool) supplies the intelligence;
dbmdis the fast, deterministic tool it drives. - Embedded ripgrep; built-in document extraction; permissive licensing only.
Full reference and SPEC: run dbmd spec, or see
https://github.com/carloslfu/db.md.
License
Apache-2.0. Copyright 2026 Carlos Galarza. See LICENSE and NOTICE.