Deposit to researchoutputs
When an AI research session finishes an analysis, it can deposit the dataset, figure, code, or document straight into the right grant — as a FAIR record with a funder-compliance check and a signed, verifiable receipt. You are the accountability layer after the workbench.
What happens when you deposit
- Deposit — an agent calls
deposit_outputwith the file (or an https URL) and its FAIR metadata (title, licence, output type) against one of the owner's grants. - Draft — the file lands private, as a draft. It does not become public and does not count toward compliance until a human reviews it.
- Publish — the grant owner reviews and publishes it on the website (the human step). Publishing mints an Ed25519-signed deposit receipt — a verifiable proof of what was deposited, by whom, and when.
- Count — the published output updates the grant's compliance and flows into the institution and funder portfolios (as verified or self-reported).
Endpoint
A Model Context Protocol server over streamable HTTP (JSON-RPC). Read tools are public; write tools (deposit_output) require your access token.
https://ldhooktwashovvmygqly.supabase.co/functions/v1/mcp
Point a client at ?space=<id> for a single-grant endpoint. Tools: search_spaces, get_space, get_citation, get_reuse, check_compliance, deposit_output.
Connector config
Generate a token under Your Spaces → Connect an agent, then drop this into your MCP client:
{
"mcpServers": {
"researchoutputs": {
"url": "https://ldhooktwashovvmygqly.supabase.co/functions/v1/mcp",
"headers": {
"Authorization": "Bearer ros_your_token",
"apikey": "eyJhbGciOiJIUzI1NiIsInR5cCI6IkpXVCJ9.eyJpc3MiOiJzdXBhYmFzZSIsInJlZiI6ImxkaG9va3R3YXNob3Z2bXlncWx5Iiwicm9sZSI6ImFub24iLCJpYXQiOjE3ODA0ODczOTUsImV4cCI6MjA5NjA2MzM5NX0.skB83Y9vyhBg9B9tAsPctygWFrpKmk5c9pBeCkme7Qs"
}
}
}
}Publishing always stays a human action. Deposits are rate-limited, size-limited, and fetched safely; every deposit records who deposited what, when, and its content hash. Full connector docs.