Roadmap
Every part of the product and the state it is actually in — live, preview, or a published contract with nothing behind it yet.
Three things run on recorded data. Several more are built and usable but running on preview numbers. The rest are contracts published so you can write the integration before the service answers. Collapsing those three states into a shipped/not-shipped binary would be a lie in one direction or the other, so all three are carried here and on the marketing site, from the same list.
How to read a status
| Status | What it means for a caller |
|---|---|
| Live | In production, on recorded data. Depend on it. |
| Preview | Built and usable in the app, but running on preview data until the recorder is wired to it. The response shape is real; the figures are not yet. Every screen that shows one says so. |
| Roadmap | Described precisely enough to build against, and not built. No endpoint answers. Write the integration if you want — just do not ship it behind a check that assumes a 200. |
The foundation
The part that cannot be caught up on later. Depth is state, not history — a competitor starting today waits for the next rotation to have a comparable record.
| Capability | Status | Where it is written down |
|---|---|---|
| Continuous depth recording | Live | Depth recording — every five minutes per pool since 22 August 2026, append-only. |
| Disclosed method | Live | Measurement method — assumed size, window and depth source on every result, with method_version. |
| Open-source verifier | Live | `@crifine/sdk` — recompute any published payload and disagree with us. Built and tested to be able to say no. |
| Keyless verification endpoint | Roadmap | Evidence — the contract every result is designed around. evidence_url is present on published results; the endpoint itself is not serving. See Verification. |
The core number
What a size actually clears at. All of it is built; all of it is waiting on the recorder for its figures, except the two Ethereum pools now read live from the chain — see what shipped on 4 September.
| Capability | Status | Where it is written down |
|---|---|---|
| Pre-trade fill estimates | Preview | `GET /v1/exit/{pool}` · Pre-trade checks for agents · estimate() in `@crifine/sdk` |
| Live pool pages | Preview | Pool pages — quoted against realized on one axis, plus the ladder and slippage by size. |
| Depth ladders | Preview | `GET /v1/ladder/{pool}` · crifine ladder in the CLI |
| Slippage forecasting | Preview | maxSizeFor() and simulateOrders() in `@crifine/sdk` · crifine size and crifine simulate |
| Spread history | Preview | `GET /v1/history/{pool}` — the recorded series, once the window is long enough to be worth quoting. |
Distribution and reach
| Capability | Status | Where it is written down |
|---|---|---|
| Public fragility board | Preview | `GET /v1/fragility` — stays free permanently. A thin book is a public fact. |
| Portfolio view | Preview | `GET /v1/portfolio` · Portfolio scoring · Dashboard |
| Spread alerting | Roadmap | `/v1/alerts` · Alerting — including the re-arm and hysteresis rules, so the contract is complete before it serves. |
| Route comparison | Roadmap | rankRoutes() in `@crifine/sdk` computes it today from ladders you supply; no hosted endpoint ranks venues yet. |
| Stress-window queries | Roadmap | GET /v1/stress/{pool} — worst depth and worst gap across a chosen period. Sketched in Liquidation exposure. The screen exists in the app on preview data; what is missing is the series behind it, not the code. |
| Depth decay tracking | Roadmap | How fast a book thins after a price move. The screen exists in the app on preview data. It needs paired move-and-depth observations across several events, and the recorder has not seen enough of them yet. |
| Multi-chain coverage | Roadmap | Recording runs across Ethereum, Base, Arbitrum and Robinhood Chain; broad venue and lending-market coverage past that is not built. |
RWA and collateral
The session fields already exist and are already applied — what is roadmap here is the scoring built on top of them, not the flag. See Market hours and RWA.
| Capability | Status | Where it is written down |
|---|---|---|
| Tokenized equity scoring | Roadmap | market_open and session_note ship on results today; execution scoring across equities and treasuries does not. |
| Market-hours risk flags | Roadmap | Weekend, after-close and holiday exposure isolated as its own signal rather than inferred from one boolean. |
| Redemption-window modeling | Roadmap | Time-to-cash on RWA assets. Nothing on-chain reports it, so it is a modelled figure and will be labelled as one. |
| Liquidation exposure scoring | Roadmap | Liquidation exposure — score collateral at cascade size against lowest_depth_7d_usd, not at the median swap. |
Machine-native delivery
| Capability | Status | Where it is written down |
|---|---|---|
| Agent-native JSON | Live | The MCP endpoint — 8 read-only tools, no account, answering today. The hosted REST surface behind the API reference is not. |
| Machine-payable endpoints | Roadmap | Payment · Paying with x402. The client half is live: `@crifine/x402` enforces maxPerCall and scoped budgets. Settlement is not. |
| Sub-second responses | Roadmap | A response-time target, not a measured one. It becomes a claim when there is a service to measure. |
| Webhooks and streams | Roadmap | Delivery into existing execution and risk stacks. Blocked behind alerting, which is blocked behind the recorder. |
| Accuracy record | Roadmap | Accuracy record — every estimate compared against what actually happened. |
What gates what
Most of this list is not waiting on engineering, and saying so is more useful than an ordered backlog would be.
- The recorder gates every preview row. They are built. They render. They are waiting for a series long enough that the number underneath is worth acting on —
days_observedis how you tell how far along that is on any given result. - Series length gates stress windows and depth decay. Worst-observed over a period is a claim about a period. There is no series before 22 August 2026 and no honest way to manufacture one.
- The accuracy record gates the $CRIF accuracy bond. You cannot collateralize a claim whose error distribution nobody has measured. That ordering is stated publicly at the accuracy record and is not negotiable for the sake of a launch.
- x402 settlement gates the paid endpoints. Verification is outside that gate on purpose and stays free permanently — see Payment.