Visivo CLI
BI as code: YAML dashboards, a semantic layer and a local MCP server through which an agent drafts charts.
- 104 ★ on the repository
- Last commit
- GPL-3.0
Extended access
Why this level
- The MCP server reads data from connected sources and takes writes without authentication on localhost.
Install
- Works in
In your terminal, with SkillFoxx CLI
npx skillfoxx add cli/visivoDetects the agents on your machine, shows the access level and pins the version.
Other ways to install
Access and checks
How we reviewExtended accessThe tool runs code or makes network calls and only reads external data.
- The MCP server reads data from connected sources and takes writes without authentication on localhost.
This is third-party code. Review the repository files before installing.
What it does
Visivo describes data sources, a semantic layer and dashboards in YAML. The project lives in git and a dashboard change goes through a branch and pull request. `visivo serve` starts a local server, rebuilds the project on every save and opens a workspace where you explore a source, draft insights against live data, lay out a dashboard by drag and drop and commit the result back to YAML. For agents the server exposes MCP at `http://localhost:8000/api/mcp/`. The `list_*`, `get_*`, `validate_*` and `write_*` tools handle sources, models, metrics, charts and dashboards, while `query_source`, `profile_model` and `preview_insight` read data. Chart advisors and the `choose-a-chart`, `design-a-dashboard` and `make-it-interactive` skills encode rules for Plotly types. The agent never writes files or commits: each write stays a draft a human reviews.
Who it is for. Analysts and data engineers who keep reports in git and want an agent to draft charts from shared metrics.
Good fit when
- A dashboard has to pass review in a pull request together with other code.
- A revenue metric is defined once and every chart computes it the same way.
- An agent builds a model, charts and a dashboard in an empty project through MCP.
- Browser filters should re-aggregate data without queries to the warehouse.
Not a fit when
- Reports are built by nontechnical colleagues who prefer clicking in a ready BI system.
- You need a Python below 3.12 without installing the binary.
- The agent must commit changes itself: the tool deliberately leaves them as drafts.
Example request
Read the project's semantic layer, add an average order value metric and draft a weekly chart with a city filter.Limitations
pip installation needs Python 3.12 or newer. The standalone binary installed with `curl -fsSL https://visivo.sh | bash` is available for macOS 15 and newer, Linux x86_64 and Windows through WSL. The MCP server on localhost runs without authentication, so its address must not be exposed outside. The built-in agent uses your Anthropic, OpenAI or Google key or inference through a Visivo Cloud account. Publishing preview stages needs Visivo Cloud, while local work runs fully without it. The project is licensed under GPL-3.0.
How to disable. Remove the entry with `claude mcp remove visivo` and stop `visivo serve`. Uninstall the package with `pip uninstall visivo`; the binary lives in `~/.visivo/bin`.
FAQ
How do I connect an agent to Visivo?
Run `visivo serve` and `claude mcp add --transport http visivo http://localhost:8000/api/mcp/`. Cursor, Gemini CLI and other MCP clients use the same address.
Can the agent break my YAML?
No, agent writes land as drafts. You review them in the workspace and commit or discard them yourself.
How do I get a working example quickly?
Run `visivo init --example ev-sales`: it creates a project, starts the server and opens the browser. There are also `github-releases` and `college-football` examples.
Do filters work without warehouse queries?
Yes, inputs filter, split and sort pre-computed insight data in the browser through DuckDB-WASM.
Where do I keep naming rules and definitions?
In the `AGENTS.md` file created by `visivo init`: the built-in agent and an external client read it before writing.
Discussion
No questions or reviews yet.
Related
CLI toolsvi-sql is a terminal SQL client with a built-in MCP server: the agent sees the schema and drafts queries straight in your editor
A CLI that turns CSV, JSON and benchmarks into interactive ECharts charts and statistics in one HTML file, for people and agents
MCP Dashboards: interactive charts in chat
MCP Dashboards
32 tools draw interactive charts, KPI widgets and dashboards right inside the chat in Claude Desktop, VS Code and other clients
FlexViz builds dashboards over 100 million rows that stay interactive, and the agent reads back what you selected.
Semiotic MCP
Semiotic
Semiotic's MCP server: an agent picks the chart type, renders SVG and validates the configuration against component schemas.
A visualization language for agents and an MCP server: neat charts from a simple semantic spec
README in short
The README shows how to get a dashboard in a minute, what sets the project apart in CI and the semantic layer, and how to connect an agent. The rest lives on the docs site.