Перейти к содержимому

He Open-Sourced a $30,000 Trading Terminal — 32,418 Stars on GitHub

The AI Index

0:00 / 0:00

He Open-Sourced a $30,000 Trading Terminal — 32,418 Stars on GitHub

6 349 просмотров · 22 ч назад
The AI Index
458 подписчиков
6 349 просмотров · 22 ч назад
A Bloomberg Terminal costs $31,980 a year, per seat. This repository is free, has 32,418 GitHub stars, and claims to replace it — so I read the whole codebase to find out what is actually in there. What I found is a real engineering project: 794 MB of C++/Qt, a genuine Model Context Protocol host, 37 AI agents. It also has a licence that changed underneath its users, and an open edition that has quietly become the second-class one. This is a repo audit, not a sponsored review. Every number is sourced at the bottom. CHAPTERS 0:00 Hook — $30k vs free 0:29 What a Bloomberg Terminal actually is 1:08 Stack reveal — not Python, it's C++/Qt 1:38 100+ data connectors (all free public feeds) 2:12 16 brokers — the India-skew origin tell 2:38 The MCP integration (the real surprise) 3:16 37 AI agents — Buffett, Graham, Munger… 3:50 Boot it — 57 screens of feature creep 4:27 Reality check — the licence, the stars, what changed 5:20 Verdict — install if / skip if THE PROJECT FinceptTerminal — https://github.com/Fincept-Corporatio... Website — https://fincept.in Licence — AGPL-3.0 Windows 10/11 x64, Linux x64, macOS 13+ Apple Silicon (no Intel Mac build) Latest release at recording — v4.5.0, 1 September 2026 WHAT I VERIFIED · 32,418 stars, 4,585 forks, 186 watchers — GitHub API, 13 Oct 2026 · 1,118 commits, 834 of them (~75%) from one author — GitHub API · 156 issues filed in the project's lifetime; 5 open — GitHub API · 56.9% C++ / 42.4% Python; repository size 794 MB — GitHub languages API · 57 screen modules including maritime, geopolitics, prediction markets — repo tree · MCP implementation: 31 files, 353 KB — up from roughly 140 KB in May 2026 · 37 named agents (Buffett, Graham, Munger + 34) — these are LLM persona prompts with different system instructions. They are not trained on those people. The engineering around them is real; the intelligence is a prompt. THE PRICE COMPARISON, HONESTLY · Bloomberg Terminal — $31,980/year single seat ($2,665/month); Vendr quotes $24,000–27,000 · Refinitiv Eikon — $22,000–24,000 · FactSet — $12,000–18,000 · S&P Capital IQ — $10,000–15,000 Sources: CostBench (last verified 10 Oct 2026) and Vendr (2026). So "$30,000" is defensible — and it is the top of the range. But you are not paying Bloomberg for a user interface. You are paying for curated data, the chat network, and an SLA. A free front end on free public data is not the same product, and any video that skips that is selling you something. FOUR THINGS THAT CHANGED AFTER THE MOST POPULAR VIDEO ON THIS REPO · The licence. In May 2026 the terms required a paid commercial licence for any business or internal use, with liquidated damages "starting at USD 50,000 per organization per year". The current README says Fincept no longer sells a separate commercial or academic licence — but docs/COMMERCIAL_LICENSE.md is still in the repository, now unlinked. Read it before you build. · The open edition was demoted. The project now ships one release a month "because the team's daily work is on Enterprise". Live broker routing and real algo trading are Enterprise-only. · The star count. The widely-quoted 21,838 is stale by about eleven thousand. · The ratio. 32,418 stars, 156 issues ever filed. That says more about how this spread than about how many people use it. SOURCES · Repo stats, commit counts, issue counts, file sizes, languages, release dates — GitHub REST API, 13 Oct 2026 · Bloomberg, Refinitiv, FactSet, S&P Capital IQ pricing — CostBench (last verified 10 Oct 2026, 15 sources, medium confidence) and Vendr (2026) · Licence terms — LICENSE (AGPL-3.0) and docs/COMMERCIAL_LICENSE.md in the repository · Launch history — the maintainer's own Show HN post, September 2024 · Product screenshots — the project's own images/ directory and fincept.in If you trade Indian retail markets, or you want a real MCP playground wired to live market data, this is worth a clone. If you need Bloomberg-grade curated data, wanted a one-command install, or have a lawyer who cares about an AGPL file sitting next to a commercial licence — skip it.