OpenData Skills
Skills teach your AI assistant how to do specific things, like pulling data from OpenData or generating a publication-ready chart. Install one once and your assistant knows how to use it from then on.
What is a skill?
Think of a skill as a small instruction file that lives alongside your AI assistant. It tells the assistant: “here's how to do X, here's when to do it, here are the rules.” The assistant reads it the same way you'd read a how-to guide, then it can do that thing whenever you ask.
For OpenData, we publish two official skills. One teaches your assistant how to find and pull data. The other teaches it how to turn that data into a chart. They're designed to work together: ask for a chart of inflation, and the first skill grabs the numbers while the second builds the visualization.
The two OpenData skills
opendata-api
Pull data from OpenData into your AI conversation. Search datasets, inspect columns, filter rows, run SQL.
openchart
Generate publication-quality charts and tables. Bar, line, scatter, sankey, and US tilemap, all from natural language.
Install both in Claude Code
Claude Code has a built-in plugin marketplace. Add ours once, then install whichever skills you want. Run these three commands in any Claude Code session:
/plugin marketplace add tryopendata/skills/plugin install opendata@opendata-skills/plugin install openchart@opendata-skills
After installing, restart Claude Code. The skills load automatically whenever you ask a question that matches what they do.
Set up your API key
The opendata-api skill needs an API key so OpenData knows who's asking. The openchart skill works without one.
- Go to Settings → API Keys and click Create API Key. Give it a name like “Claude” so you remember what it's for.
- Copy the key. It starts with
od_live_. You'll only see it once, so paste it somewhere safe. - Add it to your shell so Claude can pick it up automatically:export OPENDATA_API_KEY=od_live_your_key_here
Put that export line in your ~/.zshrc or ~/.bashrc so it sticks around between terminal sessions.
Which skill should I use?
You'll usually want both, but here's a cheat sheet for when each one kicks in:
| If you want to… | Reach for… |
|---|---|
| Find a dataset, pull rows, run a query, or export data as CSV | opendata-api |
| Turn data into a chart, table, network graph, or sankey diagram | openchart |
| Both at once (“chart US unemployment over the last 5 years”) | Both — the assistant chains them automatically |
What can I ask?
Some real examples to try after you install:
- “What datasets do you have about climate?” — uses
opendata-apito search. - “Show me the columns in fred/cpi” — inspects a dataset.
- “Pull US unemployment by month since 2020” — fetches rows.
- “Chart the result as a line chart” — hands the data to
openchart. - “Make a sankey diagram of energy flows in the US” — both skills working together.
Using these skills outside Claude Code
Each skill is a standalone folder following the Agent Skills spec. You can use them in OpenAI Codex, Gemini CLI, Cursor, VS Code, and anything else that supports the spec. Copy the skill directory into your platform's skills location (usually .agents/skills/):
# Clone the marketplace repogit clone https://github.com/tryopendata/skills.git# Copy the skill you want into your projectcp -r skills/plugins/opendata/skills/opendata-api .agents/skills/opendata-api
Check your AI tool's docs for where it discovers skills. The tryopendata/skills GitHub repo has the full source and instructions.
One more skill in the marketplace
Our marketplace also includes svg-design, a separate skill for creating SVG logos and icons. It's not OpenData-specific, but it's handy if you're putting together brand assets alongside your data work. See the repo for details.
Next Steps
- OpenData API Skill guide — examples, alternative install paths, troubleshooting.
- OpenChart Skill guide — chart types, design tips, the figure editor workflow.
- Use OpenData with AI — the simpler MCP option that doesn't require Claude Code.