DEC-0026: Initial Quota and Packaging Policy
Decision
Adopt an initial product policy for quotas, packaging, and user-facing framing across the four access tiers established in DEC-0024.
The policy goal is:
- public knowledge should stay generous
- free accounts should get a real semantic-search tool
- paid membership should feel like deeper access, not a cynical gate
- MCP should remain invite-only until demand and instrumentation justify a broader developer product
Initial Tier Policy
Public
Public visitors receive:
- KB, KG, imagery, timeline, log, and public episode/library surfaces
llms.txt,llms-full.txt, and machine-readable markdown exports- public lexical search
Public search should remain metadata-oriented and low-risk. It should not be positioned as anonymous semantic retrieval.
Free account
Free signed-in users receive:
- semantic search as a real research tool
- signed-in research workspace surfaces as they come online
- enough usage to build habit and reveal value
Initial quota target
Free semantic search should launch with a conservative but real allowance:
- target default: 25 semantic searches per day per signed-in user
That number is meant as an initial product default, not a permanent law. The system should be built so it can be revised after observing actual usage.
Paid membership
Paid membership should unlock:
- AI chat
- materially higher semantic-search quotas
- future deeper research and archive tooling
Initial quota target
Paid semantic search should be high enough that ordinary use does not feel constrained:
- target default: 250 semantic searches per day per paid user
Chat remains a paid surface. It may still carry abuse controls or soft operational limits, but not a prominently marketed "message cap" unless later evidence requires it.
Initial pricing posture
The project should package paid membership as a modest but serious upgrade:
- target launch band: $5–9/month
- working default for planning and copy: $8/month
This is a packaging default, not a permanent sacred number. The point is to give the implementation and copy pass a stable center of gravity.
Developer / MCP
MCP remains:
- invite-only
- approved-account only
- identity-bound
- not bundled automatically into paid membership by default
Paid membership alone should not imply MCP access during phase 1. Developer access remains a distinct approval path.
Pricing posture
No public self-serve MCP pricing is set yet.
The near-term policy is:
- approved access by request
- no marketplace packaging yet
- no public usage-based tariff yet
When demand justifies it, the likely future model is:
- a low-friction free developer allowance
- then usage-based pricing or a higher developer plan
But that is intentionally deferred until the system has better quota telemetry and support expectations.
Packaging Standard
Membership copy
Membership should be framed as:
- deeper semantic and AI access
- fuller research tooling
- support for keeping the core public
It should not be framed as:
- hiding the knowledge
- forcing basic understanding behind payment
- punishing ordinary visitors for browsing
Developer copy
The /for-agents surface should say plainly:
- the public knowledge layer is open
- MCP is a developer surface
- MCP is available by request in phase 1
- approved accounts receive identity-bound keys
Operational standard
The first durable quota implementation should expose:
- the configured default quotas
- whether a user is on free or paid allowances
- whether MCP access is approved
Internal ops surfaces should show the current quota policy and whether the app is still on scaffolding counters vs. durable enforcement.
Context
DEC-0024 established the constitutional split, and PR-0158 added the first quota and machine-access scaffolding. PR-0159 then provided market research:
- OpenAlex validated the "free data, paid service" split
- competitor podcasts validated that the audience will pay, but offer no real knowledge infrastructure
- MCP pricing is still early enough that invite-only access remains the correct posture for now
The missing piece was concrete policy. The repo had scaffolding but no settled defaults for semantic quotas, paid packaging, or developer separation.
Consequences
Positive
- The next implementation pass can stop guessing about quotas and pricing posture.
- Membership and developer copy can speak plainly instead of gesturing vaguely at "deeper access."
- Paid and developer tiers stay distinct instead of collapsing into one blurry premium bucket.
Tradeoffs
- The initial quotas are product hypotheses, not yet evidence-optimized.
- The system still needs durable quota storage before these numbers are fully meaningful in production.
- A later decision may revise the $8/month planning default after actual user response.
Implementation Priority
The next product-facing pass should:
- surface the new tier story in
/membershipand/for-agents - replace placeholder quota numbers in code with the DEC-0026 defaults
- expose current quota mode in internal ops views
- leave public MCP pricing unset