This hub gathers how product access is packaged and what clients forfeit. The mainline, settled 2026-02-11, is a [no-free-tier launch] (Decision_DropFreeTier) — \"No free version at launch\" — paired with hard [cancellation terms] (Decision_CancellationTerms): cancellation strips all MCP tooling, leaving only raw repo files. The driving rationale is [Tony's gaming-risk argument] (Claim_FreeTierGamingRisk) — a free tier invites abuse via read-heavy 2-writer/100-reader patterns. Both current decisions explicitly supersede earlier, more generous stances: a [permanent post-payment free tier] (PriorFreeTierPolicy) and a [read-only tier activating on full payment] (PriorDeliveryGuarantee), both replaced the same day. So the trajectory runs from payment-gated free access toward no free access plus tool-loss-on-cancel, with monetization simplicity and abuse-prevention as the stated motives.
Witness phrases
verbatim No free version at launch; prevents clients from gaming 2-writer/100-reader model. Can revisit post-PMF.
verbatim Cancellation = loss of all MCP tool access (compile, query, introspect, remember, drafts). Client keeps raw repo files only.
verbatim Clients could game a free tier by having 2 writers and 100 readers
verbatim Free license tier (read-only) only activates upon full payment.
verbatim The no-free-tier decision is explicitly reversible — "Can revisit post-PMF" — not a permanent stance.
verbatim Two superseded policies still describe a world with a free tier (permanent post-payment; read-only on full payment) that someone may reach for.
verbatim The entire no-free-tier choice rests on one abuse hypothesis (read-heavy gaming); if that risk is mitigated otherwise, the rationale weakens.
verbatim Cancellation terms keep raw repo files but revoke all tooling — a partial-retention stance distinct from full lockout or full access.