MIRA · A new medium for research

What it is

A shared structure for research in progress.

MIRA is an open architecture, still in early development, that defines a shared schema for research records and the relations that connect them. Notebook plugins, authoring tools, and publishing platforms all speak this format, so researchers using different tools collaborate on a shared graph. It is early work, built in the open with the labs and toolmakers already using it.

An interface between tools

Any tool can be extended to create, query, and write to a graph. MIRA itself hosts nothing.

Stewarded by its users

MIRA is maintained in the open by the teams that build tools around it and the researchers who use it.

Why it matters

Higher fidelity on the research you care about.

The benefits below are what we expect a shared, connected record of research to make possible. Most are early predictions from the architecture, and we are testing them now with the labs already using MIRA.

For funders

See the whole field

Query the live graph of a field you fund and see which claims are well supported, which are contested, and which questions still have no answer. See more than what's in the spotlight.

Catch dropped lines of insight

When a line of work stalls, the graph shows where it stopped: the question that never got an answer, the proposed experiment still waiting for a lab.

Spot bottlenecks and emerging trends

The shape of the record shows where a field is stuck and where momentum is building, years before citation counts catch up.

Recognize the full breadth of contributions

Credit attaches to each claim, question, and piece of evidence a researcher adds. Work that would once have vanished into a co-author line can be recognized and rewarded.

Forecast where a field is going

A provenanced record is one you can compute on. When people or agents draw new conclusions from it, every inference traces back to vetted evidence.

For labs

A shared map across labs

Collaborating labs work from a shared graph of claims, evidence, and requests. That shared map gives them the precision to take on questions too big for any single lab.

Surface your unknown unknowns

Structuring work as claims, evidence, and requests reveals what's already latent in your research: the insights you haven't named and the questions you didn't know you were asking.

Assisted peer review

Most of peer review is untangling what a paper claims and what its evidence shows. In the graph that structure is explicit, so AI can carry a real share of the work.

Granular attribution

Every record is signed by the person who made it. However many labs share a project, each researcher can point to the exact contributions that are theirs.

AI as a trusted collaborator

A graph of claims and evidence is the rules of science in machine-readable form, and agents run far better on it than on cluttered PDFs. They can navigate and add to the record while human contributions stay plainly in evidence.

What's next

Help establish the shared record for science.

The MIRA Kickoff participants standing and kneeling on a lawn in front of tall trees, smiling at the camera.

From the blog

What we built at the MIRA Kickoff

Four days in Ireland, three real user stories, six working prototypes, a schema tested against all of them, and the demo videos to show for it.

Read the report

Coming up

October 12–15A workshop at the Institute of Open Science Practices in Leiden, the Netherlands.

A new medium for research.