We gave the terminal away for two years. It worked — and then it stopped being enough. An honest account of what a free, public-data build cannot do, and what changed when we stopped trying to make it.
We released the Fincept Terminal as open source because the pricing of this industry is indefensible. A Bloomberg seat runs about $27,000 per user per year. Refinitiv is roughly $22,000, FactSet around $12,000. None of those numbers reflect what the software costs to run — they reflect what a captive market will pay.
So we shipped an open one. AI agents, 100+ data connectors, the QuantLib suite, multi-asset analytics — a native C++20 terminal, public on GitHub under AGPL-3.0, free for anyone using it for themselves, for learning or for academic research. No card, no seat, no sales call.
It worked. People used it for real decisions — independent PMs, students, small RIAs, quant teams building on top of the data. That was the point, and it is why the open-source edition is not going anywhere.
But two years in, we kept running into the same wall, and it was not an engineering wall.
Every serious feature request we got past a certain point ran into one of three problems, and none of them could be solved by writing more code.
Public data sources are good enough to show you a market. They are not good enough to build a defensible position on. The datasets that are — deeper history, faster refresh, coverage the free feeds simply do not carry — come with licences, and those licences do not permit shipping the data inside a build anybody can download and fork. That is not a policy we chose. It is the terms under which the data exists.
A single AI assistant answering a question is cheap. A research workflow that plans the work, delegates to specialist agents for equity, macro and cross-asset legs, reads live data and a stack of filings, and then writes a sourced note — that is a different order of compute, and it is consumed per query, per user.
You cannot offer that to unlimited anonymous downloads. Not "we would prefer not to" — the arithmetic does not close. Somebody pays for the tokens, and if it is not the person running the query, the feature does not ship for anyone.
Every conversation we had with a fund, a family office or a research desk arrived at the same set of questions. Where does our data live? Who can see it? Can we get an audit log? Does it do SAML? Is there an SLA?
Those are not things you sprinkle on a public build. They are architecture — separate account systems, separate storage boundaries, an authorisation edge that enforces entitlement on every request. We could not bolt them onto a codebase whose defining property was that anybody could fork and run it.
Fincept Terminal Enterprise is a separate, closed-source desktop build. It runs on proprietary data, it has real compute behind its research agents, and it carries the controls a compliance team will actually sign.
It is not the open terminal with a paywall in front of it. It is a different application with 41 modules. Some of them, like the agent tooling and the quant lab, exist in the public build too — the difference is what sits behind them:
We did not close the source of the free terminal. We did not remove features from it. We did not quietly stop maintaining it while implying otherwise.
The open-source edition is still on GitHub under AGPL-3.0, still free for personal and academic use, and everything shipped stays shipped. It also has a page of its own now setting out the licence in plain terms — including the AGPL network clause people most often miss.
What we will not do is pretend the two editions are the same product. They are not. New development happens in Enterprise. If you want the modules above, that is where they are, and you pay for them.
Seats are $99, $199 or $299 per user per month. Monthly, no annual commitment, no seat minimum. You can put one analyst on it and add the second when it has earned itself.
Against $27,000 a year for a Bloomberg seat, the top Fincept tier is about 87% cheaper. We are not claiming to replace Bloomberg for a bank — Bloomberg has the deepest historical archive in the industry and its messaging network is a genuine moat. We say so on the comparison page, including the rows where we lose.
What we are claiming is that for research depth, automation and cost per seat, a desk should look hard before signing the renewal.
The honest summary of the last two years: giving the terminal away proved the demand and built the foundation, and it could not fund what came next. The private edition is what came next.
Both are live. Both stay live. One of them is where we build now.
No. It stays public on GitHub under AGPL-3.0 and everything already released stays released. Two things did change, and we would rather state them than let you find out: new development happens in the Enterprise edition, and the public terminal now ships roughly one update a month rather than continuously. The open edition is also AGPL-3.0 rather than a permissive licence, so if you modify it and run it as a service, the copyleft terms apply. For a desk we point people at Enterprise instead.
Private and proprietary datasets the public build cannot reach, far higher API and AI usage limits, multi-agent research with planning and specialist delegation, a private dataroom for your own documents, and the institutional controls — SSO and SAML, audit logs, role-based access, data isolation and an availability SLA. It is also where active development happens: the public terminal now ships about one update a month. Several capabilities exist in both editions, notably the agent tooling and the quant lab, with far higher limits and private data behind them on Enterprise.
Per user, per month: Exclusive $99, Exclusive+ $199 and Exclusive Pro $299. Quarterly billing saves 15% and grants a whole quarter of credits up front. There is no annual commitment and no seat minimum, and the subscription lives on the account rather than the machine. A Bloomberg Terminal seat is roughly $27,000 per user per year by comparison.
Two reasons, both structural rather than commercial. The data licences behind the private datasets do not permit redistribution in a public build, and the research agents consume real compute per query that somebody has to pay for. A free edition can show you the market honestly; it cannot run a fleet of research agents against licensed data for everyone who downloads it.
Multi-agent research, a quant lab with backtesting, derivatives and macro coverage, and a private dataroom your agents read alongside the market — in one private terminal, from $99 per user, per month.