Features

Everything in one window

A tour of what mxds does. Almost all of it is in the Free plan.

Agents

Run the agent you already pay for, and see what it is doing.

✻ Claude Code ~/work/shop-analytics > Why did paid search revenue stay flat? ⏺ Bash(mxds db sql --conn shop "select channel, …") ⎿ 6 rows in 1.8 ms ⏺ Orders grew 9.8%, revenue per order fell 6.8%. The new campaign sells the $40 starter kit. ⏺ Write(q3_report.md)

Claude Code, Codex, opencode or any other CLI agent, in tabs or splits
Working, idle or waiting for you, at a glance
Past chats from all three, searchable, resumed in one click
Comment on the agent’s report or terminal output, send every note back at once
Plan limits in the status bar, and caffeine that keeps your Mac awake while an agent works

CLI and MCP

Your agent drives the app while you watch.

$ mxds db sql "select channel, count(*) from orders group by 1" rows[3]{channel,count}: Paid Search,4812 Organic,6205 Email,3118 $ mxds chart build bar channel count ok: chart opened in Results › Chart

The mxds command in every mxds terminal, MCP for any client
Opens tabs, runs queries, builds charts, edits notebooks
Answers as compact, token-optimized text

SQL editor

Thirteen databases, one editor.

select channel, count(*) as orders, round(avg(revenue), 2) as avg_order from analytics.orders where order_date >= date '2026-07-01' group by channel order by orders desc;

Completion that knows your schemas, tables and columns
Format, lint and Explain; go to the definition of a CTE or alias
GitHub Copilot suggestions, if you turn them on
Every query kept in a searchable history

Results

For you and for your agent.

channelordersrevenue

Organic6 205488 917

Paid Search4 812612 340

Social3 902196 480

Email3 118301 552

Rows stream into a local DuckDB store, so large results stay fast
Histograms, null and distinct counts for every column
Data bars, pinned columns, search, aggregates of the selection
Export to CSV, TSV, JSON, SQL, Parquet or Excel

Notebooks

Jupyter, natively.

[3]

%%sql shop_analytics select country, sum(revenue) as rev from analytics.orders group by 1

country	rev
US	742 118
DE	388 405
JP	301 990

Any .ipynb on a real kernel
SQL cells on your saved connections, results handed to Python
A variable inspector, DataFrames opened in Results
Images, tables and interactive outputs inline

dbt

Preview, build and trace any model.

Preview renders the Jinja and runs it on your target
dbt build from the editor, after seeing what it will rewrite
A lineage graph you can search, rearrange and export

Charts

15 chart types from any result.

Linex weeky revenuecolor channel

Line, bar, heatmap, funnel, waterfall and more
Built by you or by your agent

Recall

Memory that outlives the chat.

Q3 revenue: paid search AOV dropfinding · just now

orders.revenue is net of refundsconvention · 3d ago

stg_orders is incremental on updated_atgotcha · 2w ago

A knowledge base for each project, on your Mac
Local embeddings and vector search, no cloud

Git and pull requests

Watch the edits land.

models/marts/fct_revenue.sql +4 −1

- o.revenue,

+ o.revenue - coalesce(k.kit_revenue, 0) as revenue,

+ o.order_seq = 1 as is_first_order

from {{ ref('stg_orders') }} o

A live diff, revert a hunk, commit what you pick
GitHub and GitLab pull requests through your own gh and glab

Terminal

Drawn on the GPU, like the rest.

Tabs and splits
Sessions survive restarting mxds
Suggestions from your shell history as you type

Workspace

Everything else you reach for.

Every connection in one tree, passwords in the Keychain
A file browser, and docs for the table you are looking at
A usage dashboard of your queries and your agents’

Native

Rust and Metal, no Electron.

Opens in half a second, 120 fps on ProMotion
Light and dark themes that follow macOS
Apple silicon, macOS 14 or later

Try it on your own data

Download for Mac