Telumin for funds & teams

Bookless by design.

Your positions are your alpha — we don’t want them. Telumin ships evidenced signals to where your book lives, and the matching happens on your side: in your spreadsheet, your agent, or your own node.

No book upload · no broker access · signals come to you

Four ways in — ordered by how little you share.

Every fintech pitch dies at “connect your data to our cloud.” Telumin never asks — the architecture doesn’t need it. Pick the mode that matches your comfort; move only when you want more.

Mode 0 — Pure signal feed

You share: nothing. An API key. Evidenced signals arrive over REST and HMAC-signed webhooks, with tiered redaction. You match them against your book inside your own systems.

Mode 1 — Watch universe

You share: a ticker list. No sizes, no cost basis — closer to a public watchlist than to a book. Every name gets continuous per-name relevance and on-demand investigations.

Mode 2 — The agent joins in place

You share: nothing. Your analyst opens Claude with two connectors — Telumin’s threats, graph, and methodology beside your own book — and the join happens in the context window. We never see positions; your custodian never sees us. Works today, zero code.

Mode 3 — On your own node

You share: derived events only. Matching and exposure math run inside your environment; the engine ships you signals, your node does the rest. Scoped with design partners — by conversation.

Institutional memory, kept alive

The value funds pay for that retail never fully sees: the reasons behind every position, monitored — so the IC memo stops being a dead document.

Living IC memos
Every position carries its thesis and invalidation rules, checked continuously against filings, prices, and news.
Drift alerts
When reality moves away from a written thesis, the review triggers itself — not at the next quarterly.
A “why” that survived a skeptic
Three roles debate every candidate signal; what ships is the argument that survived, with sources and timestamps attached.
A decision log for DDQs
Accept or reject, each with a reason — an audit trail of portfolio discipline, ready for due-diligence questionnaires.
Two clocks on every signal
Impact horizon and decision window are separate axes — a structural threat can still demand a decision before the open.
Knowledge that outlasts turnover
When an analyst leaves, the reason the fund holds the position stays with the fund.
What arrives

Signals built for a PM’s clock, not a news feed

Not summaries — a summary of every filing is just prettier noise. Typed triggers decide what deserves investigation; a three-role debate decides what deserves your attention.

The watch never sleeps

A 5-minute scan cycle across six source families — filings, fundamentals, news, prices, social, operational events. Silence is a first-class, auditable output.

Keys, not credentials

A REST API with HMAC-signed webhooks for systems; an OAuth 2.1 MCP connector with scoped tools for agents. Never broker access, never an aggregator.

Priced where your book lives

Magnitude lands as a signed dollar figure against real positions — and in the bookless modes, that math runs on your side, not ours.

Compliance by architecture

No book upload means no client-confidentiality breach and no market-data re-licensing problem. EU residency, hard deletes, and a release-timestamp clause against front-running.

Four attributes, every signal

Direction, dollar magnitude, impact horizon, decision window. A PM needs both directions — a negative-only feed is half a product.

Modeled impact
direction − · pre-open−$1,240

Local inference

The debate engine runs on our own hardware with local models. Your universe feeds no third-party AI — and the unit economics survive your scale.

Talk to the founder

No sales team, no discovery-call maze. Write one email; you’ll hear back within one business day.

Useful to include:

  • · What you run — fund, family office, or a team managing its own book
  • · Roughly how many names you’d want watched
  • · The mode you’re leaning toward — feed, watch universe, agent, or your own node

The questions funds actually ask

Straight answers on data boundaries, delivery, and money.

Data & security

Delivery & integration

Commercials