A growing family of Model Context Protocol servers — each a cited, point-in-time data layer for a different domain. Connect the ones you need to Claude, ChatGPT, Gemini, or any MCP client.
One protocol — bring any of them to the assistant you already use
The Model Context Protocol lets your AI call these servers directly — no copy-paste, no exports.
In plain English, inside your AI client.
Claude, ChatGPT, or any MCP client picks the right calls.
Reads primary sources — SEC & SEDAR filings, 13F, funding, and more.
Point-in-time and traced to the filing it came from.
Every step is read-only and auto-approvable — nothing invented, nothing mutated. See how to connect
Each MCP is a focused, cited data layer for one domain — on its own subdomain, ready to plug into your AI client.
Cited SEC filings, 13F institutional flow, activist campaigns, adviser (ADV) data, and XBRL fundamentals — point-in-time.
Live · mcpCanadian mining research from filed technical reports — NPV, IRR, tonnages and grades, extracted from NI 43-101 filings on SEDAR and verified against the source.
Live · mcp-sedarPrivate-company funding — rounds, investors, and valuations — with SEC context where private companies cross over.
Live · mcp-fundingA self-contained corpus of material contracts and legal agreements — searchable, comparable, and cited to the clause.
Live · mcp-agreementsAnnounced and closed transactions, acquirers, terms, and comparables — with SEC context on every deal.
Live · mcp-mergersPoint-in-time drug pipelines, survivorship-complete base rates, program trajectories, and target landscapes — cited to the 10-K.
Live · mcp-bioArticle-level events with a 9-class taxonomy, a materiality score, and 3/5-day price-move correlation — deduped and cited.
Live · mcp-newsEarnings, scheduled events, and macro releases that move markets — dated, cited, and point-in-time.
Live · mcp-catalystGeneration, capacity, and the data behind the AI power-demand trade — cited and point-in-time.
Live · mcp-powerIssuer debt, bond cards, holders, and refinancing screens across the corporate-debt panel — cited to the filing.
Live · mcp-trancheGuides for how each desk uses the family — the workflow, the objections, and a prompt pack to run today.
Every server returns the same trust contract — so the answer isn't just fast, it's defensible.
What was actually knowable on a date — no hindsight. Real backtests, not leakage.
The failures are in the denominator, not just the winners the model remembers.
Every number carries its filing and snippet. NULL means unknown, never invented.
SEC filings, 13F, N-PORT, ADV, and funding rounds — live, past any training cutoff.
We test this on a 50-question finance-analyst benchmark we run internally — questions that require citing specific filings, dates, and figures. Grounded in Kaleidoscope data, the model's self-contradictions drop by half, and it answers questions it previously couldn't answer at all (0 of 50 → 10 of 50). The lift comes entirely from the data — same model, same prompts.
Request a demo and we'll map your workflow to the right MCP — or several — in your own AI client.
Request a Demo