Pricing
Start free. Pay when the record proves itself.
The public ledger is free — that is the track record. Paid tiers open with the first public scoring on October 15, 2026. Priced like independent research, not a data terminal: a newsletter sells opinion; Foreglass sells a scored record and a model you can query.
Public Ledger
The open track record.
- The full public forecast ledger — sealed, dated, scored
- Scenario library & published analyses
- LLM-vs-Foreglass scoreboard
- Structural/education datasets, each with its own recorded source and reuse terms
Ledger+
For the individual analyst.
- Everything in Public Ledger
- Extended & pre-release predictions beyond the public set
- Reaction alerts — news event → sealed simulated delta
- Full historical calibration & skill data
- Monthly forecast briefing
Desk
For the institution that runs its own questions.
- Everything in Ledger+, up to 5 seats
- Run your scenario — the queryable model + counterfactual "what-if" runs
- Foreglass API / MCP seat for your own agents
- Priority resolution requests & direct analyst access
- Quarterly scenario review
- Roadmap input as a design partner
Developers & agents
Ground your agents in a real model
Instead of an LLM's hallucinated macro number, call an accounting-consistent simulator that shows its work. A read-only MCP is planned; until it ships, the same read path is open today as plain files — see the agent quickstart — and through the foreglass Python reader on PyPI (pip install foreglass; see Developers).
Read-only MCP
A thin wrapper over the same read API — counterfactuals, country cards, calibration drift — free like the files it wraps. Not yet available: no server URL, install path or registry listing today. Until it ships, fetch the published files directly; see the agent quickstart.
Metered API
Per-call and per-seat access to scenario runs for production agents, invoiced in US dollars. Priced on usage — talk to us for volume.
Paid catalog
What paid access actually contains
Paying does not buy a better number. It buys a finer cut of the same model, an earlier window on it, and the uncertainty stated in numbers instead of implied. Every line below says how far the thing behind it has been tested, in the same three words we use everywhere else - including where the answer is "not yet".
Subscriber ledger
experimental — not yet validatedEvery sealed lane, including the cuts held back from the free tier: finer wealth and sector grain, and the lanes whose due dates sit outside the public cohort. Same seals, same grading rule, same published misses.
Scenario runs on your own question
model-conditional — read the caveatYour assumptions through the same accounting-consistent engine, with the run traceable end to end. A conditional answer under stated assumptions - not a forecast of the world.
Ultra-high-net-worth tail estimates — on request
experimental — not yet validatedModel estimates below the finest grain any statistical agency publishes. No agency series exists to grade them against, which is exactly why they are labelled this way and why they are quoted per request rather than listed as a product.
Also included, and not forecasts: the calibration and skill history behind the record, and the monthly briefing. Those are measurements of what already resolved, so they carry no validation chip - a chip describes a prediction's testing, and applying one to a measurement would be a category error.
Licence boundary. The free ledger is CC BY 4.0 and stays that way. Paid content ships in the subscriber projection under its own terms and is not licensed for redistribution. Where a lane later becomes free - crisis-hazard lanes open 90 days after resolution - CC BY attaches at that point, not before.
Reference packs are not covered by the ledger's CC BY grant. The
education datasets (household payments, corporate liquidity, freight, fiscal flows,
shock-response simulation paths, market scenarios) each carry their own recorded source and license,
downloadable alongside the data itself - most are public-domain US government statistics
reused with attribution, not CC BY. Byte verification against a committed manifest
establishes provenance, not permission. Seven sources are published in this profile: six are
determined PERMITTED for redistribution; the seventh (a locally-authored unit annotation on a
permitted Federal Reserve release, itself covering two files under the same public-domain
override) is published under an assumed-permitted-by-context basis, explicitly not an
established determination - it remains flagged as unestablished in the artifact itself, and
stays that way regardless of the assumption. The Financial Contagion dimension is withheld
entirely pending an actual rights determination on its underlying source. Two sources are
withheld with it and only one is for rights reasons - the other is the computed output of the
withheld dimension and its own redistribution is permitted. Neither absence says anything
about the standing of the seven sources published here. ForeGlass's own derived
shock-response simulation paths (from a Keen–Minsky-type model) are separate from this rights system entirely and carry no reuse
grant of their own. Check each artifact's own sources[].rights.statements field
before redistributing it; this note narrows the promise, it does not extend anyone's
rights.
How to buy it today: the same way as the tiers above - a conversation and an invoice, not a checkout. Nothing here is metered or self-serve yet.
Bigger than a subscription?
When the question is a full portfolio, an institution, or a build-or-buy, the same instrument runs it under explicit assumptions with a recommendation traceable to a run. Simulation Decision Packs from $50K, first proof in days, fixed deliverables.
Questions
Why it's priced this way
Why is the ledger free?
Because the track record is the product, and a track record only means something if anyone can check it. The free ledger is how you verify us before you pay us — sealed before the fact, scored after, misses shown like hits.
Am I buying a forecast to trade on?
No. Foreglass is honest, not alpha — an accounting-consistent model, not a statistically validated or market-beating signal. You are buying grounding, scenario analysis, and an auditable track record for your own decisions. It is research and decision support, not individualized investment advice.
Why does the Desk tier cost 40× the individual one?
Same instrument, different job. Ledger+ is one analyst reading the ledger; Desk is a team running its own sealed scenarios through the model, with analyst access and priority — the difference between reading the scoreboard and putting your own question on it.
What happens on October 15, 2026?
The first batch of near-dated predictions resolves publicly against its baselines — the first entries in the scored record. Paid tiers open then, because that is when the record starts proving itself.
How does buying work today?
Personally, for now. The tier links reach a real conversation, not a checkout: we agree on what you need, we invoice, and access typically follows within a business day. Refunds are simple — if a paid tier does not do what this page says, tell us within 30 days and the invoice is refunded in full. Your invoiced price holds for the term you paid. The free ledger needs no account and never expires. Self-serve checkout arrives when demand asks for it, not before.