A citable catalogue of scientific agents.
Agentarium is to AI agents what arXiv is to papers and Zenodo is to datasets — a public, structured, citable record of the agents scientists actually use.
What it is
Each listing is a structured record — not a PDF, not a blog post — describing one agent: who made it, what it does, what model and framework and tools it runs on, what was tested, what it does and doesn't do, and a worked example. Authors get a stable, citable URL and a concept_id that survives versioning.
The one thing to remember
Agentarium verifies format and topic, not correctness. Submissions are screened for format, topic, and safety. Scientific correctness is not verified. The registry makes agents legible and comparable; it does not vouch for their results.
Who it's for
Authors — endorsed scientists at NASA-IMPACT, UAH, and beyond — submit and maintain their agents and tools. Discoverers — other scientists, students, evaluators — browse, filter, read, cite, and install agents into their own clients. Moderators endorse new authors and triage borderline submissions.
How to use it
- Discover — no sign-in. Browse, filter by domain and transparency tier, read the full record.
- Install — pick your client on any agent page; we're honest about what each one supports. The guide walks through every install path, including the bundled-Python option that needs no MCP server.
- Cite — copy the citable URL; it resolves to a specific version and never breaks.
- Contribute — sign in with ORCID, get endorsed, and submit an agent.
Who runs it
Hosted and managed by the University of Alabama in Huntsville (UAH) as community infrastructure, extending NASA-IMPACT's akd-ext framework. Governance, policies, and the moderation handbook are public and versioned.
Contact
Governance: governance@agentarium.science. Technical issues: open an issue at the repository linked in the footer.