# researchoutputs.ai

> researchoutputs.ai is a registry for everything a research project produces.
> A grant's datasets, code, models, protocols, and publications each get a
> versioned, citable record — FAIR-checked on deposit, linked to the grant that
> funded it, and rolled up to the institution, funder, and researcher behind it.
> Records are public and machine-readable: each carries schema.org JSON-LD and a
> citation, exposes per-output metadata in standard schemas (schema.org,
> DataCite, Dublin Core, BibTeX, CodeMeta, Croissant), and is queryable and
> depositable by AI agents over a Model Context Protocol (MCP) server. Given a
> grant number and funder, the platform also assembles what is already published
> under the award (OpenAlex, DataCite, Crossref, and the full NIH RePORTER
> record for NIH awards) and computes open-access and data-sharing compliance.

## Key pages

- [Home](https://researchoutputs.ai/): what the registry is and how a record is built.
- [Explore](https://researchoutputs.ai/explore): browse and search published Grant Spaces.
- [Create a deposit](https://researchoutputs.ai/new): start a Grant Space from a grant number + funder, or by uploading files.
- [Grant Space](https://researchoutputs.ai/space/): a published record (`/space/<uuid>`) — award metadata, every output with PMID/PMCID/DOI and public-access status, computed funder-compliance, a reuse graph, a citation (text + BibTeX), and schema.org `Collection` JSON-LD with `funding`, `creator`, `sourceOrganization`, and per-output `hasPart` nodes.
- [Researcher](https://researchoutputs.ai/people/): a researcher's published records, keyed by ORCID (`/people/<orcid>`).
- [Institution](https://researchoutputs.ai/institutions/): an institution's portfolio, keyed by ROR (`/institutions/<ror>`).
- [Funder](https://researchoutputs.ai/funders/): a funder's portfolio, keyed by ROR (`/funders/<ror>`).
- [Lineage](https://researchoutputs.ai/lineage): the cross-grant lineage graph — which grants' outputs fed later work.
- [Agent & workbench connector](https://researchoutputs.ai/connect): how to connect an agent to the MCP server and deposit outputs.

## Agent access (MCP)

Model Context Protocol server (streamable HTTP, JSON-RPC over POST). Machine-readable manifest: [/.well-known/mcp.json](https://researchoutputs.ai/.well-known/mcp.json).

- Endpoint: `https://ldhooktwashovvmygqly.supabase.co/functions/v1/mcp`
- Scope to one grant by appending `?space=<space-uuid>`.
- Every request needs the project's public `apikey` header (shown, ready to copy, at [/connect](https://researchoutputs.ai/connect)). Read tools need no user token; `deposit_output` needs an `Authorization: Bearer ros_…` token, issued after a one-time human ORCID sign-in and then fully agent-usable.

Tools: `search_spaces(query, limit)`, `get_space(space_id)`, `get_citation(space_id, format)`, `get_reuse(space_id, limit)`, `check_compliance(space_id)`, and `deposit_output(space_id, …)` (token required — deposits land as a draft a human publishes).

There is also a funder-scoped read REST API (compliance-of-record feed); see [/openapi.json](https://researchoutputs.ai/openapi.json).

## How to cite

Every published Grant Space renders a plain-text and BibTeX citation. Individual outputs should be cited by their own DOI/PMID where they have one.

```
{Author}. {Title}. {Year}. researchoutputs.ai. {identifier}. {url}.
```

```bibtex
@misc{key,
  author = {…},
  title = {…},
  year = {…},
  howpublished = {researchoutputs.ai},
  note = {ark:/…},
  url = {https://researchoutputs.ai/space/<uuid>}
}
```

Programmatic: `get_citation(space_id, "text" | "bibtex")` over MCP.

## How to interpret the data

- Only `published_public` Grant Spaces are readable; drafts are private.
- Each record carries a persistent **ARK** identifier. It is provisional until a production NAAN is registered — records currently mint under `ark:/99999/`, the reserved ARK test/example NAAN, which does not resolve. Cite by the identifier shown on the record.
- A grant's attribution is **verified** when the owner is proven to be a PI on the award (ORCID or name match) or is employed at the grant's institution; otherwise it is **self-reported** — treat self-reported figures as unconfirmed.
- Compliance figures are **computed** by the platform from public metadata and human-published deposits; agent-deposited drafts do not count until a human publishes them.
- Portfolio numbers carry a sample size (n), an as-of date, and a verified vs self-reported split.

## More

- [Full reference](https://researchoutputs.ai/llms-full.txt)
- [Sitemap](https://researchoutputs.ai/sitemap.xml) · [robots.txt](https://researchoutputs.ai/robots.txt) · [OpenAPI](https://researchoutputs.ai/openapi.json) · [MCP manifest](https://researchoutputs.ai/.well-known/mcp.json)