Hand it a space or earth-science question and get back a defensible research memo plus a slide deck where every finding cites a NASA brief or a real preprint — citations appendix included. A polished, ready-to-share research memo and slide deck on your space/earth-science topic, where every claim is footnoted to the actual NASA brief or preprint it came from — nothing invented, nothing un-sourced.
A one-shot Cowork (or Claude Desktop) workflow that, given a topic, discovers the right Synorb science streams, pulls cited claims from NASA (audio briefs + agency text) and earth/space-science preprints, and produces TWO finished artifacts: (1) a structured research memo (executive summary -> key findings with inline citations -> 'what changed over 30/90 days' -> a full Sources appendix of real URLs), and (2) an exported .pptx deck of the same, with a section opener and a per-finding citation footer. Optional interactive Artifact: a timeline of findings with significance shading.
It is the Cowork 'assign a task, get a deliverable' paradigm (memo + supporting analysis + deck, end to end) aimed at a NON-finance research persona Anthropic actively courts (life sciences, science research) — proving to the partnership team that Synorb generalizes far past markets. And it attacks the exact failure case that makes generic LLM science writing untrustworthy: hallucinated or mis-attributed studies. Synorb extracts structured claims from NASA's audio briefs (a shape no text-only news API has) and from EarthArXiv/ESS-Open-Archive/arXiv preprints, every one carrying a real source_url under a never-invent-URLs contract. NASA alone produced 163 manifests in 30 days (audio + text); the preprint streams add hundreds more. A defensible cited memo is the literal antidote to the platform's biggest research liability.
No setup to learn. Claude asks a handful of normal-person questions — each with a smart default, so “you pick” always works — then builds it with live Synorb data flowing in.
Smart default: A high-activity current topic surfaced from live coverage (e.g. recent NASA Earth-science or Artemis developments) if you'd rather I pick a strong one.
Smart default: A smart general reader: rigorous and fully cited, but readable without a science degree.
Smart default: Both — the cited memo and a matching .pptx deck, so you can read it and present it.
Smart default: Mission Brief — a clean, NASA-adjacent aesthetic with section openers and a citation footer on every slide.
Long-form research-report styling — serif body, pull-quotes from preprint findings, a numbered Sources appendix. Reads like a credible institute briefing.
Clean aerospace/agency aesthetic — dark section openers, finding cards with significance chips, a per-slide source footer. Presentation-ready and authoritative.
Spare and typographic — findings, dates, and a source link each, nothing decorative. For a fast, no-frills cited rundown.
RUNTIME MCP, infra-light, one-shot deliverable. The user connects the Synorb MCP to Claude Desktop/Cowork once (Add connector -> https://mcp.synorb.com/mcp, MCP token from Synorb Connect, requested in plain English at connect and kept only in the connector). On the single task run, the agent calls synorb-stream-search to find the right universe-earth science streams, uses synorb-manifests mode=orient/count (free) to map what's covered over 30/90 days, then pulls cited manifests (compact then full bodies for the chosen findings) live through the MCP. It composes those cited claims into the memo (docx) and deck (pptx) via Claude's skills, with the Sources appendix built strictly from Synorb's returned source_urls. Output is finished files the user downloads; data was live through the MCP at build time. No backend, no keys in code.
Credential safetySynorb credentials stay server-side or local — never in client/frontend code. You only enter them when the agent asks, with credentials from synorb.com/connect.
Every key finding in the memo carries an inline citation linking to the real source_url Synorb returned (a NASA page or a preprint DOI/URL), and the memo ends with a numbered Sources appendix of those exact URLs. Each deck slide footer shows the source for its finding, and a final References slide lists them. The interactive Artifact timeline makes each finding clickable to 'read the original'. URLs come only from citations[]; the agent never fabricates a study or a link, and any web-search gap-fill line is visibly tagged '(web)' and excluded from the cited-findings count.
Yes. The build prompt connects Synorb at build time to learn the real data shape, then wires live Synorb data into the running product. A polished, ready-to-share research memo and slide deck on your space/earth-science topic, where every claim is footnoted to the actual NASA brief or preprint it came from — nothing invented, nothing un-sourced.
Only where they are safe. RUNTIME MCP, infra-light, one-shot deliverable. The user connects the Synorb MCP to Claude Desktop/Cowork once (Add connector -> https://mcp.synorb.com/mcp, MCP token from Synorb Connect, requested in plain English at connect and kept only in the connector). On the single task run, the agent calls synorb-stream-search to find the right universe-earth science streams, uses synorb-manifests mode=orient/count (free) to map what's covered over 30/90 days, then pulls cited manifests (compact then full bodies for the chosen findings) live through the MCP. It composes those cited claims into the memo (docx) and deck (pptx) via Claude's skills, with the Sources appendix built strictly from Synorb's returned source_urls. Output is finished files the user downloads; data was live through the MCP at build time. No backend, no keys in code.
Every key finding in the memo carries an inline citation linking to the real source_url Synorb returned (a NASA page or a preprint DOI/URL), and the memo ends with a numbered Sources appendix of those exact URLs. Each deck slide footer shows the source for its finding, and a final References slide lists them. The interactive Artifact timeline makes each finding clickable to 'read the original'. URLs come only from citations[]; the agent never fabricates a study or a link, and any web-search gap-fill line is visibly tagged '(web)' and excluded from the cited-findings count.
A real app. Synorb is the cited data core; it also composes the platform's own tools: Cowork agentic multi-step execution (gather -> synthesize -> produce memo + deck in one assigned task), pptx skill (generate the exported slide deck) and docx skill (the formatted memo with a real Sources appendix), Artifacts: interactive findings timeline / significance heatmap as a visual companion to the memo (inline SVG, no external libs), Native web search — GAP ONLY: to add a one-sentence layperson definition of a technical term or fill a context gap Synorb doesn't cover, clearly tagged '(web)' and never used for a headline finding, Generated section-opener imagery (optional) for the deck's title and section slides for a finished, presentation-ready look.
A few plain-English questions, each with a smart default so you can just say 'you pick': What space or earth-science question or topic should the memo cover?; Who's the audience — a general reader, a technical/scientific reader, or executives/decision-makers?; Do you want just the memo, or the memo plus a slide deck?; Pick a look for the deck and memo: Editorial, Mission Brief, or Minimal..
The same Synorb MCP connects to every major coding agent. Pick yours.
Generate credentials, connect Claude, and ship something with live, cited intelligence.