Open sourceSingle binaryNo vendor softwareNo network access

Instrument files, readable by AI agents

OpenReadout is an open-source reader for lab-instrument files, designed for AI agents. It reads 96 formats from microscopes, mass spectrometers, cytometers, electrophysiology rigs and other instruments, without the vendor's software. Your agent gets metadata, images, traces, spectra and tables as JSON, and image previews it can understand.

Install Try it in your browser

# paste into your agent's chat: it reads the skill and installs the rest
$ curl -fsSL https://raw.githubusercontent.com/openreadout/openreadout/main/skills/openreadout/SKILL.md
# or, with openreadout installed, register it yourself
$ openreadout self skill --install all
$ openreadout mcp --install claude-desktop   # or claude, cursor, codex, vscode, gemini …

# install (macOS, Linux; Windows, Homebrew, npm and more on the install page)
$ curl -fsSL https://raw.githubusercontent.com/openreadout/openreadout/main/scripts/install.sh | sh
# what is in it, is it intact, look at it, convert it
$ openreadout info cells.lif
$ openreadout check plate7/
$ openreadout preview cells.lif --composite
$ openreadout export cells.lif -o cells.ome.tiff

# wheels for Linux, macOS and Windows
$ pip install openreadout
>>> import openreadout
>>> with openreadout.File("cells.lif") as f:
...     plane = f.read_plane(image=0, c=0, z=0)   # a NumPy array

There is no packaged release yet. Install lists the other options.

Section of a whole mouse, stained, decoded from a Zeiss CZI slide scan — A whole-mouse section from a 3.7 GB Zeiss slide scan: 190,309 × 69,378 pixels at 0.22 µm. OpenReadout reads about 2.5 MB of the file to draw the overview, and only the tiles it needs for each zoom. How it works.

Zoomed view of the mouse section — A whole-mouse section from a 3.7 GB Zeiss slide scan: 190,309 × 69,378 pixels at 0.22 µm. OpenReadout reads about 2.5 MB of the file to draw the overview, and only the tiles it needs for each zoom. How it works.

What it does

InspectImages, channels, traces, spectra, tables and metadata, as text or JSON. CheckFinds truncated files, missing planes and damaged structure, and exits with code 4 when a file is corrupt. ExportOME-TIFF, OME-Zarr, mzML, NWB, CSV, Parquet, Arrow, JCAMP-DX, Allotrope ASM and RDML. Each export is read back and verified. PreviewImage planes, traces, spectra and plate heat maps as PNG, with rulers and a scale bar. AnalyzeChromatogram peaks, plate assays, qPCR, NMR peaks, patch-clamp features, spikes and flow gates, with documented methods. BatchRuns over whole folders, indexes lab shares and watches running acquisitions.

Decoded from raw files

Each panel is drawn from what OpenReadout read out of a public sample file, with no vendor software and no conversion.

For AI agents

Connect OpenReadout to Claude, Cursor, Codex, VS Code, Gemini or another agent with the skill or the MCP server, then ask about your files in plain words. The tool output below is what OpenReadout returned for public sample files.

Stable JSON: Every command takes --json and prints JSON with a published schema.
Exit codes: 0 ok, 3 unknown format, 4 corrupt, 6 unsupported feature. Agents branch on the code, not the message.
Errors with a hint: Each error says what to try next, so the agent can fix its own mistakes.
Previews: preview writes a PNG the agent can look at, so it can see the image, trace or plate it is reasoning about.
Cheap metadata: info reads headers only, so a 100 GB file costs the same as a small one. --only returns just the fields asked for.
Assurance: Each result says whether files like it were validated against an independent reader, so the agent knows when to double-check.

An agent asks for an image that does not exist.

$ openreadout preview cells.lif --image 3 --json
{
  "ok": false,
  "error": {
    "code": "usage",
    "message": "usage error: image 3 not found (file has 1 images)",
    "hint": "Indices are zero-based; `openreadout info FILE --json` lists the images, traces (sweep_count, sample_count), tables (row_count) and spectra the file holds.",
    "exit_code": 2
  }
}

The agent follows the hint, lists the images and picks the right index. Connect an assistant.

Formats

Area	Formats	Export to
Light microscopy	Zeiss CZI, Nikon ND2, Leica LIF, Olympus OIR/VSI/OIB, Imaris, OME-TIFF and other TIFF variants, OME-Zarr, whole-slide images	OME-TIFF, OME-Zarr
High-content screening	Harmony (Opera Phenix, Operetta), ImageXpress, CellVoyager	OME-Zarr plate, OME-TIFF
Electron microscopy	MRC, Gatan DM3/DM4, FEI SER/EMI, Velox EMD	OME-TIFF, OME-Zarr
Mass spectrometry	Thermo RAW, Bruker timsTOF, Agilent MassHunter, Waters MassLynx, Sciex WIFF, mzML	mzML, Parquet, Arrow
Chromatography	Agilent ChemStation and OpenLab, Shimadzu, Chromeleon, AIA/ANDI	CSV, JCAMP-DX, Parquet
Electrophysiology	Axon ABF, Intan, SpikeGLX, Open Ephys, Neuralynx, Blackrock, Plexon, HEKA, Spike2, NWB	NWB, CSV, Parquet
NMR and spectroscopy	Bruker TopSpin and OPUS, Varian, JEOL, Thermo OMNIC, Renishaw, JCAMP-DX, SPC	JCAMP-DX, CSV
Flow cytometry	FCS, FlowJo workspaces, Gating-ML	CSV, Parquet, Arrow
Plate readers and qPCR	Plate-reader exports, RDML, Applied Biosystems, LightCycler, Rotor-Gene	Allotrope ASM, RDML, CSV
Other	ÄKTA, ITC, Biacore, Seahorse, Octet, Zetasizer, XRD, EPR, electrochemistry, thermal analysis	CSV, Parquet

All 96 formats, with what each reader covers and its known gaps.

Tested against real files

Readers are tested on about 1,500 public instrument files and compared with independent libraries such as czifile, nd2, Bio-Formats, FlowIO and pyABF. Pixel data must match exactly. Validation.

Your files are read-only

OpenReadout doesn't modify your files. Exports go to a new file, which is read back and compared with the original before it gets its final name.

It works offline

OpenReadout doesn't connect to the internet, so your data stays on your computer. Readers are written from public files and permissively licensed documentation, under a clean-room policy.

Where to go next

InstallOne binary for macOS, Windows and Linux. Your first fileinfo, check, export and preview in five minutes. Connect an assistantThe skill and the MCP server for Claude, Cursor, VS Code, Codex, Gemini and others. RecipesCheck a folder, fit an IC50, index a lab share. One command each. Python and RPixels as NumPy, dask or xarray arrays, tables as data frames. How it worksA look inside instrument files, and why OpenReadout reads so little of them.

About the files. The images and examples use public files: Zenodo records, OME sample data and the test data of open-source projects such as pyABF and FlowKit, used under their licences. The damaged-folder example was made by cutting a real file short.