Coursework figure
Import a lab CSV, create named variables, build an X&Y card, set log axes if the data needs them, annotate, and export a vector PDF.
Built for
Different jobs. The same workstation.
A thesis figure due Friday, a manuscript panel that has to survive peer review, or an experiment that can't leave the building - one native Mac app covers all three. Pick the one that's yours.
The fast path
Thesis-quality figures, without the pandas detour
Import a messy lab CSV, name the variables that matter, and build an X&Y plot or a fitted histogram - without rebuilding the pipeline in pandas every time you reopen it.
Vector PDF export with embedded fonts at journal DPI is in the Free tier, so the first submission-ready figure is never behind a checkout.
Three ways in
What a study session actually looks like
Thesis draft
Save the project, reopen it weeks later, adjust the card, and export again - without hunting for the notebook cell that made the old plot.
Learning bridge
Get the analysis shape right in the app, then read the Python execution trace - the exact code Autoplot ran - when you want to understand or modify the pipeline.
FAQ
Questions before the deadline
Shouldn't I just learn Python?
Yes - and that is the point. It matters to understand the first principles of the analysis, not just the figure that comes out.
Autoplot is not a replacement for learning Python; it is a replacement for losing an afternoon to pandas documentation when the real task is a deadline figure. Every operation the assistant runs leaves a Python execution trace you can open, read, and copy - so you can study the exact code that produced your result.
Is the Free tier enough for coursework or a thesis figure?
Usually, yes. Free is permanent - not a trial - and covers:
- X&Y plots, histograms with fits, and PDF, log-binned, and power-law analysis
- Project save and load, plus vector PDF export
- A small hosted-assistant allowance
Plus opens the heavier surfaces - CDF and CCDF, scaling, heat maps, correlations - if your work grows into them.
Can I export something I can actually submit?
Yes. Autoplot exports vector PDF with embedded fonts at journal DPI, plus PNG and JPG. Annotations live with the figure and move when you restyle, so the version you hand in matches the one you tuned.
Can I reopen my work later and change a plot?
Yes. A project file preserves your imported files, variables, cards, annotations, and chart appearance. Reopen it, adjust a card, and re-export - you are editing the analysis, not reconstructing it from memory.
Is there a student discount?
Pricing is uniform at launch, with no academic discount yet - but a student discount is coming soon. In the meantime, Free is permanent and covers what most coursework needs.
I have never used an analysis app. How steep is the curve?
The app follows the steps you would take anyway: import, name variables, build a card, fit, annotate, export. When you are unsure how to express an analysis, describe it to the assistant in plain language - it proposes a structured operation and waits for you to confirm before anything runs.
What files can I import?
CSV, TSV, and text files, with a parsing preview before you commit. Inspect delimiters, headers, and skipped rows, then promote columns to named variables.
Why scientists
The scientific depth is the point
Power-law fits with MLE exponents and KS-selected x_min, log-binned distributions, CDF and CCDF, E-S scaling, heat maps, and correlation matrices - built in, not assembled by hand from a Python package each time.
Project files preserve variables, cards, workspaces, annotations, and appearance, so a result you found last month reopens exactly as you left it.
Three workflows
From a tail on a log axis to a paper panel
Heavy-tail workflow
Build X&Y and CDF/CCDF cards, inspect tails on log axes, add Xmin diagnostics, compare candidate fits, and export the figure - or copy the Python the assistant generated.
Many-variable workflow
Move from selected variables into correlation matrices or heat maps when the problem is inherently 2D or many-variable, without flattening it into one generic chart.
Figure workflow
Use Compose to assemble X&Y, histogram, CDF, heat-map, E-S scaling, chart, and correlation sources into paper-backed multi-panel boards, exported as a single figure.
FAQ
Questions from people who plot their own data
Can my collaborators reproduce this?
Yes, in two practical ways:
- The project file reopens the same workspace state in any copy of Autoplot - same variables, cards, fits, and appearance.
- Or copy the Python and Matplotlib execution trace the assistant generated and hand over an inspectable script, not a black box.
Which analysis surfaces are included?
Free covers the core; Plus opens the advanced surfaces.
- In the workspace: X&Y, Histogram, Custom Fit, PDF, AUC, CDF, CCDF, Xmin, FSS, and E-S scaling.
- As specialized tabs: Heat Map with 2D fits and contours, 3D Plot, Correlations, Charts, and Compose.
How do the power-law and heavy-tail fits work?
Power-law fits estimate the exponent by maximum likelihood with an x_min selected by KS distance, alongside log-binned distributions and CDF/CCDF views. It is the analysis you would otherwise assemble from the powerlaw package and Matplotlib - here it is a card that stays part of the saved project, with the generated Python available to inspect.
What does the assistant actually see?
Your intent and your dataset's schema - column names, types, and sizes - never the raw values. It proposes a structured operation; nothing mutates the project until you confirm. The operation then runs locally and leaves a Python execution trace, so you can see exactly what produced the result. Data privacy
Which model powers the assistant?
An open-weights model - currently QWen 3.6 35B, hosted on deepinfra.com - chosen because it does not train on your inputs. It is a narrow analysis assistant gated behind your confirmation, not a frontier chatbot you paste data into.
Coming from Igor Pro?
WaveMetrics ended Mac support at Igor Pro 10: version 9.05 is the last Mac release, and on Apple Silicon it runs only through Rosetta.
Autoplot is honestly not at feature parity - Igor's macro language and waveform model have no like-for-like replacement here. What it is: a native Mac scientific workstation built around modern fitting, plotting, reproducibility, and a conversational assistant Igor does not have. CSV and text imports work today; .ibw and .pxp import remain roadmap items.
Is it really native, or a web app in a window?
Native SwiftUI for Apple Silicon, with document-based projects that open and save like any Mac file. Numerical work runs through a persistent in-app Python engine - NumPy, SciPy, Matplotlib, pandas - so the stack underneath is the one you already trust.
Do I need Plus?
Most working scientists will.
- Free: X&Y, histograms with fits, PDF / log-binned / power-law analysis, save and load, and vector PDF.
- Plus: CDF and CCDF, Xmin diagnostics, FSS, E-S scaling, heat maps with 2D fits, 3D plots, correlation matrices, charts, and Compose.
- Pro: only for heavier assistant volume.
Local-first
Analysis that does not leave the laptop
The local toolkit - import, variables, workspaces, fits, plots, Compose, and export - runs entirely on the Mac. By default, nothing remote touches the data.
The hosted assistant is opt-in and account-based; turning it off removes no part of the non-AI feature set.
SwiftUI shell
Persistent Python engine
Local project documents
Apple Silicon focus
Three situations
When the data cannot go to the cloud
IP-sensitive experiments
Analyze tabular results without uploading files to a shared dashboard or an external AI workspace. The boundary is a single, explainable line.
Small R&D teams
Use a native Mac app rather than a per-seat BI platform when the job is scientific plotting - an experiment, a curve, a figure - not a dashboard rollout.
Auditable AI boundary
Intent plus schema may leave for the hosted assistant; raw values stay local, behind a confirmation gate. Simple enough to put in front of compliance.
FAQ
Questions compliance will ask
Will our compliance team have something concrete to review?
Yes - and by default the answer is reassuring. The local toolkit imports, transforms, fits, plots, and exports entirely on the machine; nothing remote touches the data unless you turn on the hosted assistant.
If hosted AI is off-limits in your environment, disable it and the workstation stays fully useful. The boundary is one explainable line, not a diagram nobody can follow.
What exactly leaves the machine if we enable the assistant?
Your intent and the dataset's schema - column names, types, and sizes - never the raw values. The assistant proposes a structured operation that waits behind a confirmation gate; once you approve it, it runs locally and leaves a Python execution trace you can inspect.
Can we run it fully offline and disable the AI?
Yes. The non-AI toolkit works without a network connection. Only the optional hosted assistant, account, and billing paths reach out, and disabling the assistant costs you no local features.
Do you train on our data?
No. Raw values never reach the AI layer, and the open-weights model behind the assistant does not train on inputs. There is no cloud workspace accumulating your experiments.
Is there per-seat lock-in on the data?
No. Project files are local and portable, so the analysis is never held hostage in a vendor's cloud. Autoplot is a tool you run, not a platform you rent a seat on.
Is private, on-device AI on the roadmap?
Yes. The hosted assistant on Apple Private Cloud Compute is upcoming, for on-device-grade privacy without giving up the conversational layer.
Who is Autoplot not for?
It is not a business-intelligence tool, not a notebook environment, and not a general chat AI. It is a native Mac workstation for scientific plotting and analysis. If the job is a dashboard rollout or an enterprise ML platform, this is the wrong tool - and we would rather say so.
How do we evaluate it without a six-week IT review?
Download it directly or from the Mac App Store and assess the local toolkit and the Plus feature set without any cloud onboarding. The data path needs no review unless you choose to enable the hosted assistant.
Ready when you are
Join the waitlist for the launch build and the first honest changelog.