Intelligence transparency
How the Terminal scores and sources its intelligence
We publish the contract before we ship the engine. Every external provider has a documented stance. Every score is versioned, has stated inputs and stated limitations. Nothing here is legal sign-off — these are the conservative defaults the Terminal operates from while live integrations are still gated.
Evidence first
Every score and risk verdict is backed by discrete evidence items with their own source, evaluator and confidence.
Mock-first today
Live provider integrations are gated. The Terminal renders deterministic mock payloads so you can see the contract before any data leaves a device.
Cost-aware by design
Detail surfaces show their data freshness and cost class. One intelligence object can serve many users — that is the architecture, not an afterthought.
Non-custodial
The Terminal does not hold funds, sign transactions or auto-trade. It is a research surface and a non-custodial action layer when wallet flows are added.
Providers · 11 sources tracked
Provider policy registry
For every external source the engine might consume, we publish what the source is used for, what it is not used for, whether attribution is required, and the cache stance we take. Stances default to the conservative side until a provider's terms are explicitly re-confirmed.
Provider policies
What we may use, and how
Conservative draft stance for every external data source. Not a legal review.
Show all 11 policies
dexscreener
DexScreener
Allowed use
- · Public token metadata read (price, liquidity, volume, FDV)
- · Pair-level snapshots for cached display
Prohibited
- · Resale of bulk pair feeds
- · Public mirror of DexScreener-hosted screenshots or charts
Cache stance: limited
Rate limits: Public endpoints throttle aggressively. Cache reads and back-off on 429.
Compliance: DexScreener terms require review before commercial Terminal launch. Cached values are surfaced with attribution.
Last reviewed · 2026-05-07
helius
Helius
Allowed use
- · Solana RPC + enhanced transaction parsing
- · Account state lookups for wallet enrichment
Prohibited
- · Bulk archival mirror of Helius-enriched payloads
- · Sharing raw API key with downstream consumers
Cache stance: permitted
Rate limits: Plan-bound RPM and credit budget. Reuse a single intelligence object across users to amortise cost.
Compliance: API key must remain server-side. Confirm commercial tier covers Terminal load before launch.
Last reviewed · 2026-05-07
alchemy
Alchemy
Allowed use
- · EVM RPC for Ethereum/Base/Arbitrum/Polygon enrichment
- · Token metadata + transaction receipts for wallet inspector
Prohibited
- · Public mirror of raw RPC responses
- · Sharing API key client-side
Cache stance: permitted
Rate limits: CU-budgeted. Treat compute units like a paid resource — coalesce reads.
Compliance: Confirm commercial Growth/Scale plan covers projected Terminal traffic.
Last reviewed · 2026-05-07
birdeye
Birdeye
Allowed use
- · Solana token holder distribution + holder count
- · Per-mint top holder snapshots for the holders panel
Prohibited
- · Public mirror of Birdeye-hosted holder feeds
- · Sharing API key client-side
Cache stance: limited
Rate limits: Per-key RPM budget. Cache per-mint holder responses; tighten limit when degraded.
Compliance: API key must remain server-side. Confirm commercial tier covers Terminal load before launch.
Last reviewed · 2026-05-07
jupiter
Jupiter
Allowed use
- · Solana token list + metadata read
- · Reference quotes for display (never executed)
Prohibited
- · Order placement or execution from the Terminal
- · Custodial routing
Cache stance: permitted
Rate limits: Public endpoints throttle. Cache token list reads.
Compliance: Quotes are reference only. Terminal never signs or executes — confirms non-custodial rule.
Last reviewed · 2026-05-07
geckoterminal
GeckoTerminal
Allowed use
- · Cross-chain pool metadata as a sanity-check fallback
- · Volume/liquidity snapshots for cached display
Prohibited
- · Bulk export of GeckoTerminal pool feeds
Cache stance: limited
Rate limits: Public API has tight rate limits. Use as fallback, not primary.
Compliance: CoinGecko/GeckoTerminal terms must be reviewed before commercial use.
Last reviewed · 2026-05-07
solscan
Solscan
Allowed use
- · Account + token metadata read for the inspector
- · Per-tx context lookups
Prohibited
- · Bulk mirroring of Solscan datasets
Cache stance: limited
Rate limits: Public endpoints throttle. Solscan Pro may be required at scale.
Compliance: Solscan terms reserve commercial use. Confirm Pro plan terms before launch.
Last reviewed · 2026-05-07
etherscan
Etherscan family
Allowed use
- · Address + tx context for inspector
- · Contract verification source via key
Prohibited
- · Mirror of Etherscan-hosted data without attribution
Cache stance: limited
Rate limits: Per-key RPS budget. Family includes Etherscan/Basescan/Arbiscan/PolygonScan.
Compliance: Terms-of-service explicitly call out caching and attribution. Review per-network terms before commercial use.
Last reviewed · 2026-05-07
website_meta
Project website metadata
Allowed use
- · Read of public OG/meta tags from the project's own domain
- · Display of project-published links for context
Prohibited
- · Scraping behind logins or paywalls
- · Re-publishing entire site content
Cache stance: encouraged
Rate limits: One read per refresh cycle per project. Always cache aggressively.
Compliance: Robots.txt / TOS of the target site take precedence. Do not bypass auth walls.
Last reviewed · 2026-05-07
social_meta
Public social metadata
Allowed use
- · Public account existence checks via official APIs only
- · Cached display of project-published social links
Prohibited
- · Ingesting follower lists for resale
- · Bypassing platform rate limits or login walls
Cache stance: limited
Rate limits: Each platform has its own paid tier. Defer to official APIs only.
Compliance: Each platform's developer terms apply individually. Confirm commercial tier of every social API consumed.
Last reviewed · 2026-05-07
ai_summary
AI summarisation
Allowed use
- · Summarising cached evidence into a one-line rationale
- · Synthesising signal+wallet context for the inspector
Prohibited
- · Producing trade or allocation recommendations
- · Generating verdicts unsupported by deterministic evidence
- · Hallucinating addresses, transaction hashes or balances
Cache stance: encouraged
Rate limits: Per-tier daily quota. Cache outputs aggressively to amortise cost across users.
Compliance: AI must summarise deterministic evidence only. Disclose AI involvement to the user.
Last reviewed · 2026-05-07
Scores · 12 versioned entries
Score registry
Every score the Terminal renders has a name, scale, version, purpose, inputs, output interpretation and a published list of limitations. We bump the version when the formula or interpretation moves so users see the change before they see new numbers.
Score registry
Premium scores the engine emits
Versioned contracts. Bot-emitted, Terminal-rendered. No real algorithms here yet — every score is a contract until live data lands.
Show all 12 scores
wallet_alpha · v0.1.0-draft
Wallet Alpha Score
Estimate how reliably a wallet enters positions before broad attention arrives.
Inputs
- · Entry timing vs broad attentionhelius
- · Realised PnL distributionhelius
- · Win rate on closed positionshelius
Limitations
- · Cold-start wallets with few closed positions return low-confidence scores.
- · Realised PnL underestimates wallets that hold long; pair with consistency.
- · Off-chain context (CEX deposits/withdrawals) is not captured.
Output: 0–35 weak edge · 36–60 mixed · 61–80 consistent edge · 81–100 strong edge with stable conditions.
Confidence: Higher confidence when sample size > 20 closed positions and the score is stable across two non-overlapping 30-day windows.
wallet_consistency · v0.1.0-draft
Wallet Consistency Score
Estimate how stable a wallet's strategy is across time and market regimes.
Inputs
- · Trade cadence variancehelius
- · Position size stabilityhelius
- · Regime coveragehelius
Limitations
- · Wallets that hibernate may look inconsistent without being unreliable.
- · Multi-wallet actors split the same strategy across addresses, lowering apparent consistency.
Output: 0–40 erratic · 41–70 mixed · 71–100 consistent.
Confidence: Higher confidence with > 60 days of activity and at least one regime transition observed.
copy_trade_safety · v0.1.0-draft
Copy-Trade Safety Score
Estimate how dangerous it would be to mirror this wallet without further checks. Research signal only — the Terminal never executes trades.
Inputs
- · Exit speed distributionhelius
- · Size relative to pool liquiditydexscreener
- · Bot/sniper signaturehelius
Limitations
- · Score is research-only. Mirroring trades is not safe even at high scores.
- · Cannot account for off-chain context the original wallet may have.
Output: 0–30 unsafe to mirror · 31–60 caution · 61–100 mirror only with deliberate sizing.
Confidence: Higher confidence with explicit exit-speed evidence; lower when sample size is small.
dev_trust · v0.1.0-draft
Dev Trust Score
Track the historical trustworthiness of the wallet/team that deployed a token contract.
Inputs
- · Prior deployment historyhelius
- · Metadata reusewebsite_meta
- · LP handlingdexscreener
Limitations
- · New deployers with no history default to mid-band low confidence.
- · Multi-wallet teams hide history; cluster detection is heuristic.
Output: 0–35 high concern · 36–60 mixed · 61–100 historically clean.
Confidence: Higher confidence when at least three prior deploys and on-chain LP state are observable.
token_heat · v0.1.0-draft
Token Heat Score
Composite measure of attention building around a token across smart-wallet flow, narrative, and liquidity.
Inputs
- · Smart-wallet net flowhelius
- · Liquidity change ratedexscreener
- · Narrative match strengthai_summary
Limitations
- · Heat is not a price prediction.
- · AI-derived narrative match summarises cached evidence, not real-time sentiment.
Output: 0–35 cold · 36–60 warming · 61–80 hot · 81–100 sustained interest.
Confidence: Higher confidence when multiple inputs converge in the same direction within the same window.
rug_probability · v0.1.0-draft
Rug Probability Score
Estimate the probability that a token will lose substantial value due to deployer/LP behaviour rather than market movement.
Inputs
- · LP lock statedexscreener
- · Mint/freeze authorityhelius
- · Deployer historyhelius
Limitations
- · Cannot foresee social-engineering exits or off-chain treasury misuse.
- · A low score is not a safety guarantee.
Output: 0–20 low · 21–50 elevated · 51–80 high · 81–100 critical.
Confidence: Deterministic where LP/authority state is on-chain. Drops when deployer history is unknown.
liquidity_quality · v0.1.0-draft
Liquidity Quality Score
Measure depth, distribution and stability of liquidity available to a token.
Inputs
- · Pool depthdexscreener
- · Pool concentrationdexscreener
- · LP stabilitydexscreener
Limitations
- · Does not capture off-chain CEX liquidity.
- · Cannot detect routing-only liquidity that disappears under load.
Output: 0–35 thin · 36–60 modest · 61–80 healthy · 81–100 deep.
Confidence: Deterministic from pool reads when DexScreener is available.
holder_health · v0.1.0-draft
Holder Health Score
Composite of holder count, top-N concentration and fresh-wallet share.
Inputs
- · Holder counthelius
- · Top-10 concentrationhelius
- · Fresh-wallet sharehelius
Limitations
- · Cannot detect off-chain custodians holding for many users.
- · Concentration alone is not a verdict.
Output: 0–35 fragile · 36–60 mixed · 61–80 healthy · 81–100 broadly distributed.
Confidence: Deterministic from on-chain reads. Confidence drops without fresh-wallet enrichment.
smart_money_interest · v0.1.0-draft
Smart Money Interest Score
Aggregate interest from tracked smart wallets in a token over the last N hours.
Inputs
- · Active smart walletshelius
- · Smart-wallet net USDhelius
Limitations
- · Tracked-wallet universe is curated; absence of interest is not a verdict.
- · Wallets may be coordinated; treat density bursts cautiously.
Output: 0–35 little interest · 36–60 some · 61–80 strong · 81–100 unusually strong.
Confidence: Higher confidence when the wallet sample is large and conviction is high.
launch_quality · v0.1.0-draft
Launch Quality Score
Composite read on a fresh deploy: distribution, LP setup, metadata, deployer history.
Inputs
- · Initial distributionhelius
- · LP setupdexscreener
- · Metadata qualitywebsite_meta
Limitations
- · Snapshot only — quality can degrade post-launch.
- · Cannot detect human factors like team intent.
Output: 0–35 weak · 36–60 mixed · 61–80 solid · 81–100 strong launch hygiene.
Confidence: High confidence within the first 24h after deploy when on-chain inputs are complete.
social_authenticity · v0.1.0-draft
Social Authenticity Score
Estimate whether a token's social presence reflects organic interest or coordinated amplification.
Inputs
- · Engagement distributionsocial_meta
- · Account age distributionsocial_meta
Limitations
- · Coordinated amplification can mimic organic distributions.
- · Niche genuine communities can look coordinated due to scale.
Output: 0–35 low · 36–60 mixed · 61–80 plausible · 81–100 likely organic.
Confidence: Heuristic, lower-confidence by construction. Should never be the sole basis of a verdict.
alert_confidence · v0.1.0-draft
Alert Confidence Score
Estimate how confident the engine is that an alert reflects a real, actionable event rather than noise.
Inputs
- · Input agreementhelius
- · Historical precisionhelius
Limitations
- · Alert confidence is not a price target.
- · Historical precision is sample-dependent and may be unstable for new templates.
Output: 0–35 noisy · 36–60 mixed · 61–80 reliable · 81–100 high agreement.
Confidence: Higher when multiple independent inputs agree and historical precision is known.
Scope
What this page is — and isn't
This page is the public contract. It is read-only. There is no privileged surface here, and no hidden controls. Live integrations are gated behind feature flags; the Terminal renders mock payloads until those flags flip.
Research output only. Verify before acting on any signal.