causari · huygdv · Jul 5, 2026
diff --git a/docs/AGENTIC-MONETIZATION.md b/docs/AGENTIC-MONETIZATION.md
@@ -0,0 +1,248 @@
+# Agentic Monetization Strategy
+
+## Status
+
+Strategic note / candidate architecture. This is not a commitment to adopt any one payment rail yet.
+
+## Source trigger
+
+Cloudflare announced the Monetization Gateway on 2026-07-01: a planned edge-layer product for charging for web pages, datasets, APIs, and MCP tools through x402 / HTTP 402 Payment Required.
+
+References:
+
+- Cloudflare Blog: https://blog.cloudflare.com/monetization-gateway/
+- x402 Foundation: https://x402.org/
+- x402 repository: https://github.com/coinbase/x402
+
+## Thesis
+
+Causari should treat agentic monetization as a future adapter layer, not as the product core.
+
+The core product remains:
+
+```text
+curated causal events
++ causal links
++ evidence / provenance
++ confidence scores
++ pack-level insights
++ MCP/API consumption surfaces
+```
+
+A payment rail is valuable only after Causari owns resources that agents repeatedly want to call. The right order is therefore:
+
+```text
+valuable resource → stable schema → API/MCP surface → policy/pricing adapter
+```
+
+not:
+
+```text
+payment protocol → product shape
+```
+
+## Expert review round
+
+### Founder lens
+
+The strategic signal is strong: the web is moving from human attention monetization toward machine-readable resource monetization. Causari is well positioned if it becomes a trusted source of structured causal intelligence rather than another recap/news UI.
+
+Founder implication: the moat should be the quality of causal modeling, evidence trails, and pack curation discipline. Cloudflare/x402 can become distribution and payment infrastructure, but it should not define Causari's identity.
+
+### Product owner lens
+
+Users and agents will not pay for raw JSON alone. They will pay for compressed judgment:
+
+- what happened
+- why it mattered
+- what it changed
+- what to watch next
+- how confident the system is
+- which evidence supports the conclusion
+
+This matches the existing event-pack framing in `docs/PACKS.md`. The next product milestone is to make every live pack expose a small number of high-value resource types.
+
+Recommended resource types:
+
+```text
+GET /packs/{packId}/recap/latest
+GET /packs/{packId}/events
+GET /packs/{packId}/links
+GET /packs/{packId}/insights
+GET /packs/{packId}/watchpoints
+GET /events/{eventId}/causal-chain
+POST /analyze/article
+POST /verify/claim
+```
+
+### Solution architect lens
+
+Do not bake x402 into the data package. Keep causari-data as the canonical source of truth and expose monetization at the serving layer.
+
+Proposed layers:
+
+```text
+causari-data repository
+  ↓
+pack build / validation
+  ↓
+resource server / MCP server
+  ↓
+policy engine
+  ↓
+monetization adapter
+  ↓
+Cloudflare Monetization Gateway / x402 / API key / enterprise contract
+```
+
+The policy engine should be rail-neutral:
+
+```yaml
+resources:
+  - id: pack-latest-recap
+    route: GET /packs/{packId}/recap/latest
+    access:
+      free_quota: 10/day
+      require_auth: optional
+    pricing:
+      unit: request
+      amount_usd: 0.01
+    adapters:
+      - free
+      - api_key
+      - x402
+      - enterprise
+```
+
+### Security and trust lens
+
+x402-style payments are promising, but they introduce new risks:
+
+- replay or under-bound payment proofs
+- paid-but-denied outcomes
+- unpaid-service outcomes
+- accidental leakage of request metadata to facilitators
+- unclear compliance posture across jurisdictions
+- agent overspending without a user-visible budget
+
+Causari should not ask an agent to pay unless the paid resource has:
+
+- deterministic resource identity
+- explicit price quote
+- bounded response shape
+- evidence/provenance fields where relevant
+- idempotency key for repeat calls
+- clear refund/error semantics
+- audit log entry
+
+### Market lens
+
+Cloudflare's move increases the chance that many small data/API/MCP providers can charge without building full billing systems. That is good for Causari, but it also lowers distribution barriers for competitors.
+
+The defensible wedge is not charging. The defensible wedge is that Causari's resource answers are more trustworthy, structured, agent-friendly, and causally useful than generic search or generic summarization.
+
+### User lens
+
+Human users still need a frontend. Agents need stable resources. The same pack should support both:
+
+```text
+human view: recap, timeline, graph, narrative
+agent view: compact JSON, causal chain, confidence, evidence, watchpoints
+```
+
+The frontend is not just marketing; it is the review surface that makes paid machine-readable resources legible and trustworthy.
+
+## Recommended decision
+
+Adopt a rail-neutral monetization architecture.
+
+Causari should prepare for Cloudflare Monetization Gateway / x402 by defining resource contracts and pricing metadata, but avoid a hard dependency until the gateway is generally available and real agent buyers exist.
+
+## Near-term implementation plan
+
+### Phase 1 — Resource contracts
+
+Define stable resource contracts for pack consumption:
+
+```text
+resource id
+route or MCP tool name
+input schema
+output schema
+cache policy
+freshness expectation
+evidence/provenance requirement
+pricing metadata placeholder
+```
+
+Suggested first file:
+
+```text
+docs/RESOURCE-CONTRACTS.md
+```
+
+### Phase 2 — Pack metadata
+
+Extend pack README guidance with optional monetization metadata:
+
+```yaml
+monetization:
+  status: experimental
+  recommended_unit: request
+  free_surface: recap summary
+  paid_surfaces:
+    - causal-chain
+    - verification-report
+    - watchpoint-forecast
+```
+
+This should remain metadata only. The JSON data files should not require payment-specific fields.
+
+### Phase 3 — Serving adapter
+
+Implement in the resource server or MCP server, not in this data repo:
+
+```text
+free adapter
+api-key adapter
+x402/cloudflare adapter
+enterprise-contract adapter
+```
+
+### Phase 4 — First paid experiments
+
+Best candidates:
+
+1. `verify/article` — validate a third-party article, extract claims, score evidence, produce a report.
+2. `events/{id}/causal-chain` — return a compact causal chain with evidence and confidence.
+3. `packs/{packId}/watchpoints` — return next events/questions to monitor.
+4. `packs/ai-models/recap/latest` — daily structured AI model ecosystem recap.
+
+## Design constraints
+
+- Data remains portable and license-clear.
+- Paid access must be enforced outside the canonical dataset.
+- Every paid answer should include enough provenance to justify payment.
+- Pricing must be configurable per resource, not hardcoded per pack.
+- x402 is an adapter, not the domain model.
+- Agents need spending budgets and retry-safe semantics.
+- Human review surfaces remain first-class.
+
+## Open questions
+
+1. Should pack metadata live in each pack README, a `pack.yml`, or a registry file?
+2. Should monetized resources be exposed first through MCP, REST, or both?
+3. What is the minimum paid resource that feels worth paying for: recap, causal chain, verification report, or watchpoint forecast?
+4. Should Causari support stablecoin-only payments, or wait for fiat/card facilitators to mature?
+5. How should CC-BY-SA data licensing interact with paid API access and derivative reports?
+
+## Non-goals for now
+
+- Do not implement wallet handling in this repo.
+- Do not require x402 fields in `events.json`, `links.json`, or `insights.json`.
+- Do not replace human-facing recap with agent-only APIs.
+- Do not optimize for monetization before resource quality.
+
+## Working conclusion
+
+Cloudflare's Monetization Gateway is a strong validation signal for Causari's data-pack and MCP direction. It suggests that causal intelligence can become a paid machine-readable resource. The immediate action is not to add payments; it is to make Causari resources precise enough that agents would rationally pay for them.