Deploy, query, and visualize with the help of CedarDB directly from Claude Code. Includes two agents, three skills, and a bundled MCP server for direct database access.
/plugin marketplace add cedardb/claude-plugin
/plugin install cedardb@claude-plugin
/deploy
Then:
Load two months of the NYC taxi dataset and all related static data. Design a dashboard analysing the fare distribution.
Claude will invoke the analyst to load the data and the dashboard agent to build the visualization.
- Docker — required by
/deployto run CedarDB locally (install) - Node.js — required to run the bundled MCP server (
node --versionto check) - Python 3 with
venv— required by analyst and dashboard agents
Invoked explicitly as slash commands:
| Command | What it does |
|---|---|
/deploy |
Deploy a local CedarDB instance via Docker |
/connect |
Connect to an existing local or remote CedarDB server |
/stop |
Stop dashboards and optionally the Docker container |
Invoked automatically based on what you ask for:
| Agent | Trigger | What it does |
|---|---|---|
cedardb-analyst |
"Query…", "Load…", "Show me…", "Analyze…" | Loads data, runs SQL, produces interactive HTML reports |
cedardb-dashboard |
"Build a dashboard for…" | Builds a live auto-refreshing Flask + Plotly.js dashboard |
The plugin ships a custom MCP server (mcp/) that connects directly to CedarDB using the DSN written to .cedardb by /deploy or /connect. All SQL — queries, imports, schema discovery — goes through the MCP server rather than shell commands, which means no psql dependency and no credential exposure in bash history.
.cedardb is gitignored. It is read by the MCP server on every call, so switching databases is as simple as running /connect again with a different DSN.
MCP server dependencies are installed automatically on first session start via a plugin hook — no manual npm install required.
.cedardb # DSN + label (gitignored)
cedardb-data/ # Docker volume for persistent data (gitignored)
.cedardb_work/
catalog.json # Index of all analyses and dashboards
reports/ # Named SQL + HTML report pairs
dashboard/ # Live dashboard app and static files