Trust is transparency. MoIRA runs your analysis end to end and lets you check every method, script, and file.
Ask your research question in plain English. MoIRA writes the code, runs it in a sandbox, and explains the result.
MoIRA inspects the files themselves and picks the processing lane that matches. You are never asked to classify your own data.
Say what the groups are and what you are controlling for. The plan is built around your design and shown before anything runs.
R and Python written for your design, run in an isolated sandbox, with live status on every step and a stop you can hit.
Every figure, table and script is yours to download, and the conversation that produced them stays with the project.
Three domains are in production today, each validated end to end before it shipped. The rest are in the order we intend to build them.
Infinium arrays, bisulfite or enzymatic sequencing, long-read native calls, or a beta matrix you already have.
SeSAMedirect importBismarkmodkitlimmaDMRcatemethylKitDSSmissMethylmethylclockDroplet, full-length or long-read chemistries — or an existing count matrix or AnnData object.
scanpySTARsoloSTAR + featureCountsBLAZE + IsoQuantscanpyCellTypistSingleRedgeRHarmonyscVIdecouplerA gene count matrix is the default entry and needs no samplesheet at all. Raw reads are supported either way.
default entryTrim Galore + STARPychopper + IsoQuantDESeq2edgeRlimma-voomclusterProfilerfgseadecouplerONT direct-RNA needs GPU basecalling that Fargate does not provide, and PacBio inputs are recognised but not yet routed. Both are gated rather than advertised.
A project holds its conversations, its files and its storage in one place. Every preprocessing run and every analysis is filed automatically, so the project is the record rather than someone’s laptop.
Invite people as members or viewers. They open the whole conversation that produced a result, not just the figure at the end.
Any member can continue a colleague’s analysis in their own editable branch. It reads the original files in place, so it costs no extra storage.
Hand out a view-only link to someone with no account. Revoke it whenever you like; every link issued stays on the record.
The project page: who is on it, what they have run, what it costs against the project’s storage, and the view-only link — all in one place.
Every figure can be traced back through the code that produced it to the question you asked, and everything in that chain is downloadable. A result you cannot check is a result you cannot publish — so here is one, opened all the way up.
limma on M-values with your covariates in the design, as generated R — in a sandbox that refuses network modules and reads and writes only its own staged files..R or .py script that produced them — all typed and downloadable.-v2, -v3, each linked to what it supersedes. Recolouring a plot does not.No. A processed matrix is a first-class starting point in every domain — a beta matrix for methylation, a count matrix or AnnData object for single-cell, and for bulk RNA-seq a gene count matrix is the default entry and needs no samplesheet at all. MoIRA records what has already been done to the data so it does not repeat or contradict it.
Preprocessing and analysis are both gated on your explicit approval. When MoIRA proposes a pipeline it sets a confirmation flag, and the tools that would start a run refuse while that flag is set. It has to end its turn and wait for you.
Your data is encrypted in transit and at rest, analyses run in isolated sandboxes with no network access, and every query is filtered by your user. Nothing you upload is used to train models.
Runs consume credits metered by compute, and your balance is always visible in the app. Pricing is still being finalised ahead of general availability. Waitlist members will see it before anyone is asked to pay.
MoIRA is in closed beta with a small group of research labs. Join the waitlist and we will reach out as capacity opens.
Join the waitlist