SciCope

Version 0.4.2 · 18 September 2026 · Mac, Windows, Android, iPhone

Run the microscope from one app. Then find the answer in the pictures.

SciCope drives the hardware on your bench — cameras, a motorised stage, lights, a focus lens — from your Mac or PC, and analyses what it records in a workspace built for image analysis: find your objects with one slider, track them through the frames, correct the tracker by clicking, measure in micrometres and export movies, montages and tables. An assistant runs the experiment and writes new tools when you ask in plain words. Any phone can be the camera.

Phone camera apps: Android APK · iPhone app installed directly by the lab for now · all downloads and first-launch notes · no hardware yet? The demo microscope inside the app shows everything on a simulated plate.

The SciCope desktop app in dark mode: the Microscope view with the live picture of the demo microscope, the camera, stage and light panels on the right and the assistant panel.

How it works

From a bench full of cables to a number you can publish

Two halves, one app: the half that drives your hardware, and the half that turns the pictures it records into measurements. An assistant sits across both.

Connect the microscope

Open Microscope › Devices… and pick what is on the bench: a Thorlabs, Andor or USB camera, your phone, or the computer's own camera; the stage board (Marlin or GRBL) on its serial port; the light relay; the Optotune focus lens. Nothing on the bench? Open the demo microscope and try everything on a simulated plate.

Record the experiment

Jog the stage, autofocus, snap. Scan a plate in a grid or a spiral, run a timelapse, follow a moving animal, or save the whole thing as a loop program you can repeat. Or say it: “scan a 3×3 grid, then a timelapse every 30 seconds” becomes steps you see as a card — the ones that move the stage wait for your Allow.

Analyse it — and correct it

Open the session in the Analyze workspace and work like a photo editor: tools on the left, the picture big in the middle, panels on the right. Find your objects with one slider, track them through the frames, and fix what the tracker got wrong by clicking — including the moment two animals became one blob. Then export overlay movies, montages, projections and tables in real units.

Hardware & analysis

The two halves of an experiment, in one app

The first half drives the cameras, the stage, the lights and the lens on your bench. The second turns what they recorded into measurements — and lets you correct the machine where it is wrong.

The Microscope view in light mode: the live picture with position, focus score and scale bar, one toolbar with Snap, Autofocus, Timelapse, Scan, Track and STOP, and the camera settings.
The live picture as large as the screen allows, with the stage, light and lens controls folded beside it.

The whole bench in one window

Cameras, a motorised stage, the lights and a focus lens, driven from your own computer over USB and serial — no vendor software open on the side.

  • Cameras: Thorlabs uc480, Andor (SDK2), USB cameras, the computer's own camera, and phones running the SciCope app — on the same Wi-Fi or through the relay.
  • Stage, light, lens: an XZ stage with a rotating plate on Marlin or GRBL boards (jog with the keyboard, go to a position, home, zero), relay lines as switches with a pulse, and an Optotune focus lens with autofocus — once or continuous.
  • Several cameras at once, each with its own live tile; Snap all takes one picture from every open camera at the same moment and the files share a number, so they stay paired.
  • Real units from the start: µm per pixel from the optics, from two clicked points, or by moving the stage a known distance — every later measurement is in micrometres because of it.
  • One red STOP (and the Escape key) halts every move and every running job.
Two cameras streaming at once, one tile each, with Snap all for a synchronized picture and the stage jog pad beside them.
Two cameras streaming at once — a synchronized capture takes one picture from each.

Experiments that run themselves

The routines a bench actually needs, with the steps taken from the camera's own field of view.

  • Plate scans in a grid or a spiral, overlapping by the percentage you choose, and stitched back into one picture afterwards.
  • Timelapses at any interval — sub-second ones save frames from the live video — optionally at a list of coordinates, with the real frame times written next to the photos.
  • Tracking on the stage: the microscope follows a moving animal and keeps it centred; a 3D focus scan records a stack.
  • Loop programs: chain the steps — move, focus, photograph every camera, wait — save them and run them again tomorrow.
  • Pictures where you expect them: session folders with captures.csv, numbered files with a prefix and a tag, JPG, PNG or 16-bit TIFF.
The Analyze workspace: 60 frames of a worm plate, every animal outlined and named, a measurement of 5.25 mm across the plate, a 2000 µm scale bar, the Detect panel with threshold and size sliders reading “34 spots on this frame”, the object list and the film strip.
Sixty frames of a plate: every animal outlined, named and followed, with a 5.25 mm measurement and a scale bar.

An analysis workspace, not a dialog

The lab's own analysis program, rebuilt as a window you can work in: tools on the left, the picture big in the middle, panels on the right, the frames along the bottom.

  • Find your objects with one slider: dark or bright, a threshold and a size, live spot count under your thumb — or click one object with the Wand and the settings follow it.
  • Track them: automatically through all frames, or seed the ones you care about and let them be followed; each object keeps its name, its colour and its path.
  • Measure in real units: lines and boxes in µm and mm, speeds, distances and areas per object, calibration from a known length.
  • Marks that stay readable: the outline of the object instead of a big circle, a halo under every mark and label, and a focus mode that shows one animal alone.
  • Export what the paper needs: overlay movies with a scale bar and a clock, path pictures, montages, projections, filtered copies, an image calculator, cluster density on white/blue pairs, stitched plates, stage path movies, and CSV tables of every detection and every track.
The Check panel listing what needs a human: two objects touching with a “Swap from here” button, missing frames, and “One blob looks like 3 animals on frames 1–4 — worm 37 sits on it” with a “Separate into 3” button.
The Check panel: every confusion the tracker knows about, each with the button that fixes it.

When the tracker is confused, it says so

Automatic tracking goes wrong where animals meet. Instead of leaving you to find those moments, SciCope lists them and hands you the fix.

  • “One blob looks like 2 animals on frames 10–12.” A detection far bigger than one animal is a clump — the app says so, and Separate cuts it in two and gives each half back to the right animal, so the names do not swap.
  • “Worm 7 walks out of worm 3 at frame 12.” When a second animal appears on top of another, the pair were one blob earlier — the app works out since when, and can separate those frames backwards.
  • Gaps, jumps and near-misses are listed the same way; clicking a line jumps to that frame with the animals selected.
  • Correct it by clicking: point at where the animal really is and it moves there, the detection learns from those pixels, and that one animal is re-tracked from that frame on. Swap two names, split a track, merge two, follow again.
  • Your corrections stick: every fix becomes a fixed point that later automatic passes keep.
The Tool view of the desktop app: a tool named “Switch Arduino LED on and off” with a form of five settings, a Run button, a preview table, and the assistant conversation on the right offering to install a helper program.
A tool the assistant built for a lab device: a form with plain-language settings, the conversation beside it.

An assistant across both halves

It sees which cameras are open, where the stage is, what is running and which pictures are loaded — and it can act on all of it.

  • Runs the microscope: “focus, then photograph a 4×4 grid with 10 % overlap” becomes a card of steps. Motion, focus changes and series wait for your Allow; a switch in Settings lets it run hands-off.
  • Drives the analysis: it can open a folder, set the detection, track, measure and export from the same conversation — every button in the Analyze workspace is something it can press.
  • Writes new tools: a form, not a script, tested on one of your own pictures before you see it; Developer mode shows the Python behind it.
  • Reports back honestly: file names, positions and errors go back into the conversation, so its next answer is about what actually happened.
The SciCope phone app on Android: the live camera picture with lens buttons, an ISO, exposure and sharpness readout, grid, scale and level overlays, and the brightness control set to Automatic.
The phone app: lens, brightness, focus and colour controls, a sharpness readout and overlays.

Phones are cameras too

Put a phone in the eyepiece adapter and it appears as a camera in the Microscope view — the desktop app takes the real, full-resolution photo when it captures.

  • Pairing like a remote-desktop tool: the phone shows a 9-digit SciCope ID, the computer asks, the owner taps Allow once or Always allow. Direct on the same Wi-Fi, through the SciCope relay from anywhere else.
  • Lens choice, zoom, manual or locked exposure, focus and white balance; Freeze keeps the brightness fixed across a series.
  • The phone can work alone: timelapses at any interval, focus stacking, a live sharpness readout, and session folders with a captures.csv that records the scale and exposure of every photo.

Try it without hardware

The demo microscope is a simulated worm plate that pans with the stage, sharpens with the lens and brightens with the exposure. Autofocus, scans, tracking, the assistant's steps and the whole Analyze workspace work on it, so you can learn the app before the first cable is plugged in.

Pictures from any microscope

The Analyze workspace opens a folder of pictures — yours or ones another microscope's software exported. Sequences of JPG, PNG and TIFF (including 16-bit), every frame at its own time from captures.csv, or a plain numbered series; a detections CSV can be imported back.

Python built in, Fiji optional

A private Python with NumPy, SciPy, scikit-image, OpenCV, pandas, matplotlib, the camera and relay driver libraries and the lab's own analysis library ships inside the installer and sets itself up in seconds. Nothing is installed system-wide. Fiji (ImageJ) is an optional download for macro tools and microscope file formats.

Your photos, your folders

Pictures are saved on your computer in session folders you choose, and results are written next to them — one folder per tool and date, with the tables and the output pictures. With Google Drive for Desktop a phone's sessions arrive as folders too. There is no SciCope cloud.

Runs on Mac and Windows

Macs (Apple Silicon and Intel) and 64-bit Windows PCs, with light and dark themes, a command palette (⌘⇧P / Ctrl+Shift+P) over folders, tools and commands, and every conversation saved for later. The lab cameras and the NI relay need their vendor drivers, which are Windows-only.

Your own AI account

The assistant runs on the AI you choose: Claude, ChatGPT, Gemini or DeepSeek with an API key, or the Claude Code, Codex and Gemini CLI apps with your subscription. Keys are stored encrypted on your computer; there is no SciCope account.

The Microscope view in dark mode with the live picture of the demo microscope and the stage, light and lens panels.
The Microscope view: the hardware half
The Analyze workspace with a worm plate, every animal outlined and named.
The Analyze workspace: the other half
The desktop app on Windows 11, showing a folder of HEIC photos and the assistant panel.
Windows 11

Privacy

Your microscope and your photos stay yours

The engine that drives the hardware runs on your own computer, and your pictures never pass through anything we run. Here is exactly what travels where.

Cameras, stage, lightsYour computer

Hardware. Cameras, the stage board, the relay and the lens are driven by the engine inside the app, over USB and serial on your computer. The assistant's steps are shown to you before they run; nothing about your hardware leaves the computer unless you ask the assistant a question.

PhoneYour Google DriveYour computer

Photos. Photos never travel through the relay. A phone uploads them to your own Google Drive if you switch that on, and Google Drive for Desktop brings them to the computer; on the same Wi-Fi the desktop app copies them directly.

PhoneSciCope relay / CloudflareYour computer

From elsewhere the phone's control messages and live preview go through our relay, or through Cloudflare when the video needs a hand. The relay forwards traffic only after the phone's owner has accepted, keeps nothing, and each leg is encrypted. On the same network it is never used.

  • You are in charge of the stage. The assistant's steps that move the stage, change the focus or start a series wait for your Allow. The hands-off switch in Settings is off until you turn it on.
  • The phone's owner is in charge of the phone. Nothing works without Allow on the phone, or the unattended-access password you set yourself.
  • The assistant uses your AI provider. Your request, the microscope's state in a few lines, and — when it needs to look — a downsized copy of one sample photo go to the provider you connected with your own key or account. Nothing goes to us.
  • No analytics. This website loads nothing from third parties and runs no analytics.

Downloads

SciCope 0.4.2

Released 18 September 2026. New in this version: the Analyze workspace — the lab's own analysis program rebuilt as an interactive window, with object outlines that stay readable, a focus mode, and a Check panel that finds the moments where the tracking needs a human, including the blob that is really two animals. Python and every library the app needs are inside the installer, so the first start needs no download; Fiji is optional and can be added later from Settings. The desktop app and the Android app are early builds and are not code-signed yet — see the first-launch notes below.

Mac · Apple Silicon

SciCope for Mac (M1 and later)

Apple Silicon Macs, macOS 11 or newer. Disk image with Python inside, about 360 MB.

SciCope-0.4.2-arm64.dmg

Download for Apple Silicon

Mac · Intel

SciCope for Mac (Intel)

Intel Macs, macOS 11 or newer. Disk image with Python inside, about 360 MB.

SciCope-0.4.2-x64.dmg

Download for Intel Macs

Windows

SciCope for Windows

64-bit Windows (tested on Windows 11). Installer with Python inside, about 360 MB; installs for the current user, no administrator needed.

SciCope-0.4.2-win-x64.exe

Download for Windows

Android

SciCope for Android

Android 8.0 and later. The phone camera app; installs from the APK, about 69 MB.

SciCope-android.apk

Download the APK

iPhone

SciCope for iPhone

iOS 17 and later. The phone camera app.

Installed directly by the lab for now. There is no App Store listing yet.

Checksums of the published files: SHA256SUMS.txt. Google Drive for Desktop (from Google) is only needed if you want a phone's photos to arrive as folders on the computer.

First launch

What to expect the first time

The builds are not signed with a developer certificate yet, so macOS and Windows warn you once. Here is how to get past the warning and what the app does at its first start.

MacOpen Anyway

  1. Open the disk image and drag SciCope into Applications. Open it once — macOS says it cannot verify the developer; click Done.
  2. Go to System Settings › Privacy & Security, scroll down to the message about SciCope and click Open Anyway. Confirm with your password.
  3. When macOS asks whether SciCope may find devices on the local network, allow it — that is how the app finds your phone on the Wi-Fi. It asks for the camera only if you choose the Mac's own camera as a microscope camera.
  4. Python sets itself up in a few seconds on the first start, from the files inside the app. Fiji is optional, under Settings › Advanced.
  5. Open the Microscope view. Real hardware is set up under Devices… (camera, stage board and port, relay, lens); Open demo microscope shows everything on a simulated plate without any hardware.

WindowsRun anyway

  1. Run the installer. When SmartScreen says Windows protected your PC, click More info, then Run anyway. The app installs for your user account; no administrator password is needed.
  2. When Windows Defender Firewall asks whether SciCope may communicate on the network, click Allow. A dismissed prompt leaves block rules that silently stop the direct live video from a phone (fix it under Windows Defender Firewall › Allowed apps).
  3. Python sets itself up in about a minute on the first start, from the files inside the app. Fiji is optional, under Settings › Advanced. HEIC photos from phones need the HEIF Image Extensions and HEVC Video Extensions from the Microsoft Store; the app tells you if they are missing.
  4. Open the Microscope view and set up the hardware under Devices…: Thorlabs and Andor cameras need their vendor drivers (ThorCam, Andor SDK) installed on the PC, the NI relay needs NI-DAQmx, the stage board shows up as a COM port. Without hardware, Open demo microscope.

PhoneAndroid and iPhone

  1. Download the APK on the phone and open it. Android asks to allow installs from this source; allow it, then tap Install.
  2. Give the app camera access. In Settings, switch on remote control to see the phone's SciCope ID; sign in to Google Drive only if you want the sessions synced.
  3. Connect the phone from the desktop app (Microscope view → camera list → Connect a phone…, or the Phone camera section) and keep the phone app open on its Camera screen while a computer controls it; leaving the app stops remote control.

On iPhone the app is installed directly by the lab for now; the steps after installation are the same.

FAQ

Questions we hear first

Which hardware works?

Cameras: Thorlabs uc480 (DCx) and Andor SDK2 cameras through their vendor drivers on Windows, USB cameras, the computer's own camera, and phones with the SciCope app. Stages: Marlin boards (Creality-style controllers, the Ender S1 Plus build) and GRBL boards (Arduino with a CNC shield) on a serial port — X and Z plus a rotating plate. Light: NI-DAQmx digital outputs (relays) on Windows. Focus: Optotune lens drivers. The Devices dialog lists the ports and cameras it can see.

Can I use it without a stage, a lens or a phone?

Yes. Any single camera works on its own — pictures, timelapses, the analysis tools and the assistant. The demo microscope simulates a camera, a stage, a light and a lens together, so every feature can be tried with nothing plugged in.

Does the assistant move the stage by itself?

Only after you allow it. Its steps appear as a card; the ones that move the stage, change the focus or start a series wait for Allow, while taking a picture, reading values or opening a view run at once. If you prefer hands-off operation, switch on Let the assistant run the microscope without asking in Settings. The red STOP button and the Escape key stop every move and job at any time.

What happens when two animals touch and the tracker mixes them up?

It tells you. After tracking, the Check panel lists every moment that needs a human: two objects touching, a missing stretch, an impossible jump, and the one that used to ruin a dataset silently — a blob far bigger than one animal. The app compares each detection with the typical size of a single animal in your own pictures and says “one blob looks like 2 animals on frames 10–12”, or “worm 7 walks out of worm 3 at frame 12 — there were 2 animals in that spot from frame 5”. Separate then cuts the blob into worm-sized parts and gives each part back to the right animal, seeded by where each one was heading, so the names do not swap. You can also point at where an animal really is, swap two names, split or merge tracks, and re-track one animal from that frame on. Every correction becomes a fixed point that the automatic passes keep. Two animals lying exactly on top of each other cannot be told apart by any shape-based method — the app shows them sharing a point rather than inventing an answer.

Can I analyse pictures that were not taken with SciCope?

Yes. The Analyze workspace opens any folder of pictures — JPG, PNG or TIFF, including 16-bit — numbered in a sequence, with an optional filter so you can pick every second frame or one channel of an interleaved series. Tell it the scale (µm per pixel) and the time between frames, or calibrate by clicking two points on something of a known length, and every measurement comes out in micrometres and seconds. A detections CSV from an earlier run can be imported back.

Do I need Fiji or ImageJ?

No. Detection, tracking, measurement and every export run on the Python inside the app and on the lab's own analysis library — the same code that has been used on the bench for years, rebuilt as a workspace. Fiji is an optional download from Settings › Advanced for tools written as ImageJ macros and for opening microscope file formats such as CZI or ND2 through Bio-Formats.

Do I need to know Python?

No. You describe what you want in plain words, and the tool you get is a form with settings in plain words too. If you are curious, Developer mode shows the exact script that runs, and you can keep and reuse every tool. Python and its scientific libraries come with the app; nothing else is installed on the computer.

Which AI does the assistant use, and what does it cost?

Your own. In Settings you connect Claude, ChatGPT, Gemini or DeepSeek with an API key, or the Claude Code, Codex or Gemini CLI apps with the subscription you already have. SciCope has no account or subscription of its own; the AI provider bills you directly according to its own terms.

Where do my pictures and results go?

Pictures go into the session folder shown in the Microscope view, numbered with a prefix and, for synchronized captures, the camera's name. Results go into a SciCope results folder inside the session folder, one subfolder per tool and date: a summary table (CSV) with one row per photo, the output pictures and per-photo detail tables. If the folder is in your Google Drive, everything syncs.

Can I use it without Google Drive?

Yes. Pictures taken by the desktop app are saved on the computer. Point the Photos view at any folder with Change in the sidebar — including files exported by other microscope software. On the same Wi-Fi the desktop app can also copy a phone's current folder straight to the computer.

Which phones and computers work?

Phone apps: Android 8.0 and later, and iPhone with iOS 17 and later. Desktop app: Macs (Apple Silicon and Intel) and 64-bit Windows PCs. The Thorlabs and Andor cameras and the NI relay need their Windows drivers; stages, lenses, USB cameras and phones work on both systems.

Why does my computer warn me about the app?

The builds are not signed with an Apple Developer ID or a Windows code-signing certificate yet, so macOS shows the “cannot verify the developer” dialog and Windows shows SmartScreen. Both can be passed once as described under First launch. Signed, self-updating builds are planned.

How do I turn a timelapse into a movie?

On the desktop, the Analyze workspace and the Make a movie tool build an MP4 from a session folder, using the real frame times from captures.csv; the Microscope view can also record the live picture directly. The phone app can do the same from its Library.

What does the relay see?

Only what is needed to control a phone from another network: the commands, the camera settings and the live preview, and only after the phone's owner has accepted the computer. Photos and results never pass through it. Lab cameras, stages, lights and lenses are never on the network at all.