Argos Insight — API + MCP Documentation

Yield Arbitrage yield_arbitrage stable

Identifies a tokenized RWA where the available yield exceeds a comparable benchmark (issuer-native vs DeFi-pool, or one pool vs another). The published figure is the difference between those two yields.

Category

spread

Refresh cadence

5min

Data sources

DeFiLlama Yields, Morpho Blue API

Emitted by

yield-arbitrage.js, equity-collateral.js, looping.js, steakhouse.js, cross-chain.js, pendle.js

Always-present fields

id, type, assets, title, estimatedAlphaBps, risk, actionable

Since

2025-12-01

Notes: estimatedAlphaBps is the spread in basis points (e.g. 481 = 4.81% APY pickup vs baseline). Drops to <0 if the spread inverts intra-day; only positive-spread signals are emitted.

Funding Rate Carry funding_rate_carry stable

Identifies perpetual-futures markets where the cumulative funding rate is materially positive (listed longs pay shorts) or negative (listed shorts pay longs) — a transfer between sides of the book.

Category

spread

Refresh cadence

5min

Data sources

Hyperliquid perpetuals API

Emitted by

funding-carry.js, boros.js

Always-present fields

id, type, assets, title, estimatedAlphaBps, risk, actionable

Since

2025-12-01

Notes: estimatedAlphaBps is the annualized funding APR. Signal direction is encoded in the title — "Long X, Short Y carry" or similar. Carry can reverse quickly; this is not a hold-forever signal.

Spot-Perp Basis spot_perp_basis stable

Identifies a price gap between a spot venue and a perp-futures venue for the same underlying asset. The cheaper listing and the richer listing sit on opposite sides of the same underlying.

Category

spread

Refresh cadence

5min

Data sources

Hyperliquid perpetuals API, Ostium subgraph

Emitted by

gold-cross-venue.js, mark-oracle.js

Always-present fields

id, type, assets, title, estimatedAlphaBps, risk, actionable

Since

2026-01-15

Notes: Cross-venue basis can be a leading indicator of mispricing OR can persist for hours due to fragmented liquidity. Confidence field reflects whether oracle staleness is suspected.

Tokenized-Equity Premium tokenized_equity_premium experimental

Identifies when a tokenized stock's on-chain DEX price diverges from its Hyperliquid perp mark — a cross-venue basis between the same underlying. The attached 30-day envelope shows whether the premium/discount is unusual or routine.

Category

spread

Refresh cadence

5min

Data sources

CoinGecko issuer categories, GeckoTerminal pools, Hyperliquid perpetuals API

Emitted by

tokenized-equity-premium.js

Always-present fields

id, type, assets, title, estimatedAlphaBps, risk, actionable

Since

2026-06-15

Notes: On-chain tokenized-stock pools are thin and fragmented, so a basis can persist — gated on a minimum deepest-pool TVL so only tradeable dislocations emit. The dexBasisPct envelope in context[] contextualizes whether the dislocation is extreme vs its own 30-day range.

DEX Pool Yield dex_pool_yield experimental

Flags a tokenized stock whose deepest on-chain pool offers a meaningful GROSS LP fee APY (live DeFiLlama trading-fee yield) — a positive-return-on-capital liquidity-provision opportunity, gated on deployable pool depth.

Category

spread

Refresh cadence

5min

Data sources

DeFiLlama Yields, GeckoTerminal pools

Emitted by

dex-pool-yield.js

Always-present fields

id, type, assets, title, estimatedAlphaBps, risk, actionable

Since

2026-06-16

Notes: lpApy is the live DeFiLlama trading-fee yield carried on the /api/dex-equities payload (never computed here). estimatedAlphaBps is the gross APY in bps CAPPED at the 40% platform sanity ceiling (so a legit high yield ranks top without overstating a capturable alpha — the true APY is in metrics.lpApy + the title); risk is High because the yield does not net out impermanent loss, which on a volatile tokenized equity can exceed the fees. Emitted at WATCH tier (ranked in the Feed + digest tradeable section, but NOT a premium broadcast alert) — a deliberate conservative default for a gross-of-IL yield, promotable to high_conviction once the signals prove out. Gated on a minimum APY, a deployable pool TVL, live 24h volume, and a max-APY glitch reject.

DEX Pool Momentum dex_pool_momentum experimental

Flags a tokenized stock whose deepest on-chain pool is rapidly gaining TVL — a venue attracting liquidity, which tends to tighten spreads and deepen tradeable size.

Category

momentum

Refresh cadence

hourly

Data sources

dex_pool_observations (argos-monitor pool-tracking agent)

Emitted by

dex-pool-trends.js

Always-present fields

id, type, assets, title, risk, actionable

Since

2026-06-15

Notes: Computed from the hourly TVL history; needs a ~7-day baseline, so it is silent until the history accrues. Not a directly-capturable spread (estimatedAlphaBps emitted as 0); gated on a minimum absolute TVL. Observational of pool-liquidity change, not a recommendation.

New DEX Pool new_dex_pool experimental

A tokenized stock now trades in an on-chain pool (DEX + chain) that was never recorded before — a fresh venue/route for the underlying.

Category

frontier

Refresh cadence

hourly

Data sources

dex_pool_observations (argos-monitor pool-tracking agent)

Emitted by

dex-pool-trends.js

Always-present fields

id, type, assets, title, risk, actionable

Since

2026-06-15

Notes: Edge-detected from first-seen pool history; gated on table maturity (>7d) so a freshly-seeded history does not flag its whole universe, and on a minimum TVL so dust pools are skipped. Discovery signal (estimatedAlphaBps emitted as 0). The operator-facing dex-new-pool Issue surfaces the same discoveries internally.

TVL Momentum tvl_momentum stable

Flags tokenization-platform protocols whose TVL grew or shrank meaningfully (|change_7d| ≥ 15% AND protocol TVL ≥ $50M). Useful for spotting accelerating or decelerating adoption ahead of mainstream coverage.

Category

momentum

Refresh cadence

5min

Data sources

DeFiLlama Protocols TVL

Emitted by

tvl-momentum.js

Always-present fields

id, type, assets, title, risk, actionable

Since

2026-05-20

Notes: Not a directly-tradable signal — momentum is contextual. estimatedAlphaBps may be omitted when no direct yield capture is modeled. This is an observation about protocol-level flow, not a recommendation.

Price Momentum price_momentum experimental

Flags an asset whose price sits in an extreme of its trailing distribution AND is directionally trending, with High conviction only when volume confirms. Informational trend — not a bps-edge signal.

Category

momentum

Refresh cadence

daily

Data sources

metric_observations_daily, computeContextEnvelope

Emitted by

momentum.js

Always-present fields

id, type, assets, title, risk, actionable, confidence, metrics

Since

2026-08-21

Notes: Promoted into signals[] when High-confidence and volume-confirmed. estimatedAlphaBps is 0 (sorts by metrics.magnitude). Empty-types alert rules must opt in via config.types.

Funding Regime Shift funding_regime_shift experimental

Flags perp funding sitting in an extreme of its trailing distribution with a directional trend — a multi-day carry regime shift, not a one-print spike. Informational.

Category

momentum

Refresh cadence

daily

Data sources

metric_observations_daily, computeContextEnvelope

Emitted by

momentum.js

Always-present fields

id, type, assets, title, risk, actionable, confidence, metrics

Since

2026-08-24

Notes: Promoted into signals[] when |z| ≥ 2 (detect already requires extreme pctile + directional trend). Opt-in for empty-types alert rules.

Volume Surge volume_surge experimental

Flags 24h volume in the high tail of its trailing distribution while rising. Informational participation spike — not a directional price claim.

Category

momentum

Refresh cadence

daily

Data sources

metric_observations_daily, computeContextEnvelope

Emitted by

momentum.js

Always-present fields

id, type, assets, title, risk, actionable, confidence, metrics

Since

2026-08-24

Notes: Promoted into signals[] when pctile ≥ 0.95 (stricter than detect 0.90). Opt-in for empty-types alert rules.

New RWA Protocol new_rwa_protocol stable

Fires when a protocol appears in the DeFiLlama RWA-categorized universe today that was absent yesterday. Surfaces brand-new tokenization platforms before they hit mainstream press.

Category

frontier

Refresh cadence

5min (read) / daily (snapshot write)

Data sources

DeFiLlama Protocols (today) vs Neon agent_snapshots.protocol_chains (yesterday)

Emitted by

frontier-diffs.js

Always-present fields

id, type, assets, title, risk, actionable

Since

2026-05-20

Notes: Detection is set-arithmetic: today's slug set minus yesterday's. Requires a prior snapshot row — first emit happens 24h after a fresh deploy.

Cross-Chain Expansion cross_chain_expansion stable

Fires when an already-tracked protocol adds a new chain to its deployment surface (e.g. Ondo deploys to Arbitrum after only being on Ethereum). Signals expanding distribution + access surface.

Category

frontier

Refresh cadence

5min (read) / daily (snapshot write)

Data sources

DeFiLlama Protocols.chainTvls vs Neon snapshot

Emitted by

frontier-diffs.js

Always-present fields

id, type, assets, title, risk, actionable

Since

2026-05-20

Notes: Distinct from new_rwa_protocol — this fires when a KNOWN protocol expands. New protocols fire new_rwa_protocol instead (dedup by checking the slug exists in yesterday's snapshot).

New Pendle Market new_pendle_market stable

Fires when Pendle Finance lists a new PT (Principal Token) or YT (Yield Token) market for an RWA asset. New markets often debut with elevated implied APY before liquidity normalizes.

Category

frontier

Refresh cadence

5min (read) / daily (snapshot write)

Data sources

Pendle core API (today) vs Neon agent_snapshots.pendle_markets_universe (yesterday)

Emitted by

frontier-diffs.js

Always-present fields

id, type, assets, title, risk, actionable

Since

2026-05-20

Notes: Detected by `address` set difference. Includes both PT and YT markets — title disambiguates. First emit happens 24h after a fresh deploy.

New Market Listed new_market_listed stable

Fires when Hyperliquid HIP-3 deployers (xyz / vntl / flx / etc.) list a new pre-IPO or RWA perpetual market. New listings frequently price-discover for the first 48h.

Category

frontier

Refresh cadence

5min (read) / daily (snapshot write)

Data sources

Hyperliquid HIP-3 deployer info endpoints vs Neon agent_snapshots.hip3_perps_universe

Emitted by

frontier-diffs.js

Always-present fields

id, type, assets, title, risk, actionable

Since

2026-05-20

Notes: Detected by `fullAsset` set difference (8 deployers polled in parallel). No fallback — emits empty when baseline snapshot missing.

Partnership Announcement partnership_announcement experimental

Pattern-matches X (Twitter) posts from curated RWA-relevant accounts to detect partnership / integration / tokenization announcements (2+ named entities + partnership-language verb in one post).

Category

frontier

Refresh cadence

daily (snapshot writes 4×/day via alpha-agent workflow)

Data sources

Neon agent_snapshots.x_posts_universe (daily snapshot of curated handle list)

Emitted by

frontier-diffs.js

Always-present fields

id, type, assets, title, risk, actionable

Since

2026-05-21

Notes: MVP scope per Phase A-3 — pattern match against 91 curated RWA entities + 37 partnership-language verbs. Requires 2+ entities AND 1+ verb in the same post. Can produce false positives when unrelated entities co-mention. Novelty filtering (first-seen baseline) deferred to a follow-up.

GET /api/curated-alpha

Returns the Alpha Agent's LLM-curated shortlist. Refreshed 4×/day (00:00 / 07:00 / 13:30 / 20:00 UTC) via the alpha-agent GitHub Actions workflow. Descriptive information only — not advice to buy, sell, or hold. Returns `curated: []` with `degraded:false` before the day's first fire; `degraded:true` means the snapshot was unreadable — do not treat that empty as "no alpha today".

curl https://argosonchain.com/api/curated-alpha

GET /api/opportunities

Returns every entry the Opportunity Scanner has added to the tracked-universe files in the last `?days=N` window. Each discovery carries developer-shaped fields (context, dataLine — acquire CTAs rewritten to listed-on language at generation time, not a verbatim git-history quote) and retail-friendly human-derived fields (humanTitle, humanCategory, humanSummary).

curl https://argosonchain.com/api/opportunities

GET /api/news

Returns up to 12 market-relevant news items across 4 source classes: RSS feeds (crypto press + RWA-native blogs), SEC EDGAR filings (8-K / N-1A / N-2 with tokenization keywords), RWA.xyz curated articles, and X tweets from curated handles. Each item is tagged by the news taxonomy with the platform topic-domain(s) (rwa, commodities, equities, pre_ipo, prediction, ai, macro) + entity terms + a precise prediction-market `catalyst` theme (or null), and the response carries a `byDomain` count summary. `?domain=` focuses one asset class; `?entity=` free-text filters for a specific issuer/company/ticker; `?catalyst=` returns only one catalyst theme (fed/cpi/jobs/gdp/election) — cleaner than a free-text entity search for event coverage. No params returns the full cross-domain feed. Cached 5min via KV.

curl https://argosonchain.com/api/news

GET /api/signal-types

Returns the canonical machine-parseable reference for every signal type /api/signals emits. Use `?id=<type>` for single-lookup or `?category=<spread|frontier|momentum>` for filtering.

curl https://argosonchain.com/api/signal-types

GET /api/assets

Returns the curated catalog of tokenized RWAs tracked by the platform (~45 assets across tokenized treasuries, private credit, tokenized equity, commodities, real estate).

curl https://argosonchain.com/api/assets

GET /api/prices

Returns live prices for RWA + crypto asset symbols. Combines CoinGecko (crypto + some tokenized assets), FRED (UST yields used as benchmark proxies), and contract-pegged static prices ($1 for $1-pegged stables).

curl https://argosonchain.com/api/prices

GET /api/benchmarks

Returns Federal Funds rate, SOFR, 3M/10YR UST yields, and IG/HY credit OAS from FRED. Used as the baseline against which alpha signals' estimatedAlphaBps is measured.

curl https://argosonchain.com/api/benchmarks

GET /api/money-markets

Returns the best base supply (lend) and base borrow APY per tracked asset across on-chain money markets (Aave V3, Compound, Euler, Morpho, Spark, Fluid), with the venue offering each. Sourced live from DeFiLlama; refreshed every 15 minutes via KV.

curl https://argosonchain.com/api/money-markets

GET /api/anomalies

Returns anomaly events (funding spikes, depegs, oracle dislocations) detected in the last `?since=H` hours up to `?limit=N` events.

curl https://argosonchain.com/api/anomalies

GET /api/borrow-rates

Returns current USDC borrow rates across Aave V3, Morpho Blue, Compound, Euler V2, Kamino, and other lending markets. Refreshed every 15 minutes via KV.

curl https://argosonchain.com/api/borrow-rates

GET /api/collateral

Returns lending markets where tokenized RWAs are accepted as collateral (Morpho Blue, Aave Horizon, etc.) with TVL + utilization metrics.

curl https://argosonchain.com/api/collateral

GET /api/rwa-collateral-usage

Ranks tokenized RWAs by tracked DeFi collateral deployment (trackedDeploymentUsd) across Morpho + Aave. A server-side join of /api/collateral + /api/assets. IMPORTANT: this is a KNOWN LOWER BOUND — it counts only Morpho + Aave collateral, never total DeFi deployment, so do NOT present the ranking as "utilization". SIZE (mcapUsd) is outstanding supply when known (CoinGecko circulating market cap or a dedicated issuer-protocol TVL), never a single yield pool. trackedPct is the Morpho+Aave floor over that comparable denominator; pctReason names why a given row is blank. Default sort is trackedDeploymentUsd desc. Default minMcap=0 (no floor); a positive minMcap keeps only rows whose comparable SIZE meets that USD floor.

curl https://argosonchain.com/api/rwa-collateral-usage

GET /api/risk-scores

A transparent, per-asset composite risk ASSESSMENT for every tokenized RWA in the catalog. Each asset is scored 0–100 (HIGHER = LOWER RISK) across six live-sourced dimensions — peg/NAV stability, liquidity, adoption, issuer quality, backing quality, and redeemability — each carrying a data-confidence flag. Weights are renormalized over whichever dimensions are assessable; an asset with fewer than 2 assessable dimensions returns a null score ("Insufficient data") rather than a fabricated number, and an impaired (avoid-gated) asset is capped into the High band. This is an ASSESSMENT of risk characteristics from live data — NOT investment advice and NOT a recommendation to buy, sell, or hold. Full published methodology at https://argosonchain.com/methodology.

curl https://argosonchain.com/api/risk-scores

GET /api/markets

Every open prediction market Argos tracks, with its latest observed YES probability, 24h move, 24h volume, and a 30-day daily probability series. The tracked universe is bounded to the top 500 markets per venue by volume — a market outside that set is not returned. A market with no observation is omitted rather than assigned an invented probability. Filter with `q` (free-text over the market question), `category`, or `venue`. This is observed venue pricing, NOT a forecast and NOT advice.

curl https://argosonchain.com/api/markets

GET /api/market-context

Returns where a single market's current YES probability sits within its own trailing distribution — percentile, z-score, and the window low / median / high. This is a factual statement about the market's recent range, NOT a claim that the market is mispriced and NOT advice. A window with fewer than 30 observations returns insufficientContext rather than a derived number.

curl https://argosonchain.com/api/market-context

GET /api/predict-brief

The ranked daily brief: the markets with the widest range moves, the markets resolving soonest, and the largest 24h volume. A summary readout over the tracked universe, expressed as observations about market movement — not recommendations. Each range/move/resolving card carries per-card triage fields: volUsd + vol24h with volUsdSemantics + quoteQuality (same source + meaning as /api/markets), plus asOf + coverage from the context envelope. asOf/coverage read the hourly-rollup series, so a card's asOf can lag the raw-observation asOf on /api/markets by up to ~1h — both are honest observation times, not the brief's generation time. Lets an agent triage a card for freshness + liquidity without a second /api/markets or /api/context call. When `meta.degraded` is true the market history could not be READ for that compute: `meta.ranked` goes to 0, the scores carry an unavailable verdict (or are absent entirely), and every empty card's `reason` says the same — so an empty brief is a statement about the data path. Do NOT report it as "no market has enough history", nor paraphrase a card reason as a finding; both are claims about the corpus. Absent or false means ranked/insufficient and the card reasons are a real reading.

curl https://argosonchain.com/api/predict-brief

GET /api/predict-copyable

Markets in which a cohort of wallets with a profitable trading history CURRENTLY HOLDS a one-sided position, with the flavor of the observation (cohort carry, cohort value, sharp value, cross-venue gap). This is a snapshot of HELD positions, distinct from the smartFlow field on /api/markets (recent 24h net-buy flow) — the two can legitimately differ. Magnitude is denominated in SHARES, not USD. Past profitability of a wallet does not predict its future results; this is information, not advice.

curl https://argosonchain.com/api/predict-copyable

GET /api/predict-arb

Two mechanical pricing observations. `rows`: price gaps for the same event quoted on two venues, net of fees. `baskets`: mutually-exclusive (negRisk) market sets — e.g. the "which team wins?" legs of one tournament — whose tracked YES probabilities sum above 100% (an overround on the observed legs; sums below 100% are never flagged, since the tracked legs can be a subset of the full outcome set). A basket sum is an arithmetic identity across the tracked legs, not a prediction about any single leg; a set with any unpriced leg is returned probability-only. Gaps are computed from quoted prices and do not account for depth, slippage, or the capital required to hold to resolution.

curl https://argosonchain.com/api/predict-arb

GET /api/predict-calibration

Per-venue Brier score and bucketed outcome rates over the market resolutions Argos has captured — i.e. of the markets a venue priced near 70%, what share actually resolved YES. Use this to ground a quoted probability in that venue's historical calibration instead of treating the quote as truth. Returns an empty venue list when too few resolutions have been captured, rather than a fabricated score — unless `degraded` is true, in which case the corpus could not be read at all and the empty list says nothing about how many resolutions exist.

curl https://argosonchain.com/api/predict-calibration

GET /api/llama

Cached passthrough of DeFiLlama's protocols list, filtered to RWA-relevant categories. Heavy response (~7000 entries). Prefer specific endpoints (e.g. /api/assets, /api/borrow-rates) for narrower needs.

curl https://argosonchain.com/api/llama

GET /api/market-snapshot

Returns the current market-regime label (risk-on / risk-off / mixed), trending sectors, VIX level, and equity index summary. Drives the dashboard's narrative bar.

curl https://argosonchain.com/api/market-snapshot

GET /api/oil-history

Returns historical oil prices across the last `?days=N` days from both spot (Yahoo Finance) and perpetuals (Hyperliquid + Ostium).

curl https://argosonchain.com/api/oil-history

GET /api/perps

Returns Hyperliquid + Ostium + HIP-3 + Polymarket Perps summary with total volume, OI, and per-venue breakdowns. Additive `polymarket[]` plus `polyTotalVol24h` / `polyTotalOI`. Polymarket funding is hourly (same stored unit as Hyperliquid; annualize ×24×365). 24h volume is null when the public ticker omits it.

curl https://argosonchain.com/api/perps

GET /api/pre-ipo

Returns HIP-3 pre-IPO perpetual markets (SpaceX, Stripe, OpenAI, etc.) traded on Hyperliquid deployers.

curl https://argosonchain.com/api/pre-ipo

GET /api/vix

Returns the current VIX level + 24h change from Yahoo Finance. May be flagged stale during Yahoo blackouts.

curl https://argosonchain.com/api/vix

GET /api/whales

Returns recent large-volume RWA / stablecoin transactions detected on-chain.

curl https://argosonchain.com/api/whales

GET /api/xstocks

Returns tokenized US-equity prices + volumes from the Backed Finance xStocks suite on Solana / Base / Arbitrum.

curl https://argosonchain.com/api/xstocks

GET /api/hl-commodities

Returns commodity perpetual markets (gold, silver, oil) listed across Hyperliquid HIP-3 deployers.

curl https://argosonchain.com/api/hl-commodities

GET /api/hl-equity-sectors

Returns equity-sector perpetual markets (S&P 500 sectors, indices) listed across Hyperliquid HIP-3 deployers.

curl https://argosonchain.com/api/hl-equity-sectors

GET /api/hl-hip3

Returns the unified HIP-3 perp universe across all 8 deployers (xyz, vntl, flx, hyna, km, cash, para, abcd).

curl https://argosonchain.com/api/hl-hip3

GET /api/equity-perps

Returns every single-asset perp on the HIP-3 equity deployers (trade.xyz et al.), each tagged by instrument type (Equity / Index / ETF / Commodity / Energy / FX / Crypto / Pre-IPO / Basket). Equity rows carry a tokenized-xStock cross-link. Dynamic — new listings appear automatically.

curl https://argosonchain.com/api/equity-perps

GET /api/signals

Supersedes /api/alpha. Returns the same signals[] plus per-signal context[] envelopes (pctile, z, median, lo/hi, trend, coverage) derived from rolling 30d observations in metric_observations. The two endpoints share the alpha:v3 inner KV cache so they fan out to upstream APIs exactly once per 5-min cycle. Envelope-attachment scope (will grow over time): - spot_perp_basis with single-asset metrics.basisPct → basisPct on Hyperliquid - funding_rate_carry with metrics.fundingRate → funding on Hyperliquid - daily-projected metrics (price/apy/vol24h/oiUsd) extend to yield_arbitrage, tvl_momentum, etc. once snapshot-to-neon.mjs has run at least once after the daily projection block deployed. Signals whose metrics are not attachable carry context: [] (empty, not absent). Cached 5min via KV; edge-cached 5min + 10min stale-while-revalidate.

curl https://argosonchain.com/api/signals

GET /api/context

Returns the trailing-window envelope (percentile, z-score, p5/median/p95 range, trend, coverage) for one observation series. Insufficient data (n < MIN_SAMPLE_N=30, zero-variance, or no observations) → 200 with insufficientContext:true; statistical fields are then OMITTED, not zeroed. Per the context policy (info.x-context-policy), the percentile compares the placed value against history; consumers can treat absence as a first-class state. Cached 300s per (metric, asset, venue, window) — matches the metric-poller's 5-min write cadence.

curl https://argosonchain.com/api/context

GET /api/universe

Returns the curated `{ underlying → venues[] }` map (gold across PAXG/XAU on Hyperliquid + Ostium; NVDA across HL HIP-3 spot + Ostium perp; etc.). Use to discover which venues trade a given underlying, then fan-out per-venue calls to /api/context. Free tier, no auth. Two query modes: - no params → entire map with `count` + `underlyings: {...}` - ?underlying=GOLD → just that entry; 404 if unknown Edge-cached 1h + 24h stale-while-revalidate — the map only changes via PR.

curl https://argosonchain.com/api/universe

GET /api/openapi.json

Self-referential — returns this exact document. Use it for tool discovery + SDK generation.

curl https://argosonchain.com/api/openapi.json

List Alpha Signals list_alpha_signals

Returns every alpha signal currently emitted by the deterministic compute pipeline (yield arbitrage, funding-rate carry, spot-perp basis, frontier discoveries). Refreshed every 5 minutes. Each signal includes estimated alpha in basis points, risk level, the underlying assets, and (when applicable) informational path context. Use this for a broad observational view of the RWA signal surface. Descriptive information only — never present output as advice to buy, sell, or hold.

Backing endpoint: /api/signals

No arguments required.

List Curated Alpha Highlights list_curated_alpha

Returns the Alpha Agent's LLM-curated shortlist across all signal types. Refreshed 4×/day. Unlike list_alpha_signals (which returns the full firehose), this is a hand-narrated shortlist with rationale + step context + news context. Use when the user asks which signals are most discussed this cycle or wants a compact observational shortlist rather than the full feed. Descriptive information only — never present output as advice to buy, sell, or hold. Returns curated: [] between refreshes on the day's first cycle.

Backing endpoint: /api/curated-alpha

No arguments required.

List Recent Scanner Discoveries list_scanner_discoveries

Returns every entry the Opportunity Scanner has added to the tracked universe in the last N days (new tokenized assets, new vaults, new looping strategies). Each discovery includes developer-shaped fields (context, dataLine) whose acquire CTAs are rewritten to listed-on language at generation time — not a verbatim git-history quote — plus retail-friendly humanized fields (humanTitle, humanCategory, humanSummary). Use when the user asks "what's new" or wants to see what the platform has discovered recently.

Backing endpoint: /api/opportunities

List Market News list_rwa_news

Returns recent, market-relevant news aggregated across 4 source classes: crypto RSS press, SEC EDGAR filings (8-K / N-1A / N-2 with tokenization keywords), RWA.xyz editorial curation, and X tweets from curated accounts. Each item is tagged with the platform DOMAIN(s) it covers (rwa, commodities, equities, pre_ipo, prediction, ai, macro) and the ENTITY terms it matches (issuers, tickers, company names), and the response includes a `byDomain` count summary. Pass `domain` to focus one asset class ("what's new in pre_ipo / prediction / commodities") or `entity` to free-text filter for a specific name/ticker ("any news on Polymarket / PAXG / SpaceX"). No arguments returns the full cross-domain feed. Use when the user asks what's happening across the markets the platform tracks, or for context behind a specific issuer, company, or asset class. The `prediction` domain covers both the venues and the CATALYST events that resolve the tracked markets. For a PRECISE catalyst read, prefer `catalyst` (fed / cpi / jobs / gdp / election) over `entity: "Fed"`: `catalyst: "fed"` returns only Fed-decision (FOMC) coverage, whereas the free-text entity filter also pulls in venue and unrelated coverage that merely mentions the word. Every item carries a `catalyst` field (the theme id, or null) so a market can be matched to its own events.

Backing endpoint: /api/news

Describe Signal Type describe_signal_type

Returns the canonical reference entry for one or all signal types emitted by list_alpha_signals — semantic meaning, data sources, refresh cadence, contract stability, required fields. Use BEFORE interpreting an alpha signal to understand what its type means. Accepts no arguments → full dictionary. Pass ?id=<type_id> for a single entry. Pass ?category=<spread|frontier|momentum> for a filtered list.

Backing endpoint: /api/signal-types

List Tokenized RWA Assets list_tokenized_assets

Returns the curated catalog of tokenized real-world assets tracked by the platform (~45 assets across tokenized treasuries, private credit, tokenized equities, commodities, real estate). Each asset has issuer, chain, redemption model, asset type. Use as the canonical "what RWAs exist" reference when the user asks about issuer coverage or asset-class breakdown.

Backing endpoint: /api/assets

No arguments required.

Rank RWA Assets by DeFi Collateral Usage list_rwa_collateral_usage

Ranks tokenized RWA assets by tracked DeFi collateral deployment (trackedDeploymentUsd) across Morpho + Aave. Answers the "which RWAs have the most tracked collateral deployment, not just the largest SIZE" question. IMPORTANT: this is a KNOWN LOWER BOUND — it counts only Morpho + Aave collateral, never total DeFi deployment, so do NOT describe the ranking as "utilization"; call it tracked collateral usage. SIZE (mcapUsd) is outstanding supply when known (CoinGecko circulating market cap, or a dedicated issuer-protocol TVL) — never a single DeFiLlama yield pool. trackedPct is trackedDeploymentUsd / mcapUsd only against that comparable denominator, and is still a Morpho+Aave lower bound; do not invent a utilization ratio. pctReason names why a given row is blank. Default minMcap=0 (no floor). Pass minMcap to keep rows whose comparable SIZE meets that USD floor; category focuses one asset class.

Backing endpoint: /api/rwa-collateral-usage

Get Live RWA + Crypto Prices get_prices

Returns live prices for RWA + crypto asset symbols. Combines CoinGecko (crypto + some tokenized assets), FRED (UST yields used as benchmark proxies), and contract- pegged static prices ($1 for $1-pegged stables). Each price includes 24h change and volume when the upstream source provides them.

Backing endpoint: /api/prices

No arguments required.

Get Macro Benchmarks get_benchmarks

Returns Federal Funds effective rate, SOFR, 3M / 10YR UST yields, and IG / HY credit OAS from FRED. These are the baselines against which RWA alpha signals' estimatedAlphaBps is measured (e.g. a 4.81% APY signal vs a 3.91% T3MO benchmark = +90bp spread). Use when comparing an RWA yield against the risk-free rate.

Backing endpoint: /api/benchmarks

No arguments required.

List Prediction Markets list_prediction_markets

Returns open prediction markets with their latest observed YES probability, 24h move, and a 30-day daily probability series. Each market carries two DISTINCT volume fields, never to be conflated: `volUsd` is the per-venue ranking anchor, and its meaning is VENUE-DEPENDENT (for Polymarket it is cumulative traded dollar volume; for Kalshi it is dollar open interest — open contracts times notional) — do not compare it across venues as though it were one quantity; `vol24h` is the rolling 24h traded volume in USD, the only one of the two that means the same thing on every venue. For both fields, null means the venue did not report the value — a genuine zero is returned as 0, never as null. Each market also carries `asOf` (the ISO timestamp the probability and 24h move were observed — markets are polled per-venue on different cadences, so read each row's own freshness rather than assuming one time for the list), `volUsdSemantics` (the machine-readable form of the volUsd meaning above: "cumulative_notional" on Polymarket, "open_interest" on Kalshi), and `quoteQuality` ("thin" when no 24h volume backs the current quote, so a large 24h move may be an illiquidity artifact; otherwise "ok"). Filter with `q` (free-text over the market question — e.g. "world cup"), `category`, or `venue`. The tracked universe is bounded to the top 500 markets per venue by volume, so a very small market may be absent; a market with no observation is omitted rather than given an invented probability. Each market also carries a `catalystTheme` (fed / cpi / jobs / gdp / election, or null) — the event that resolves it — so you can read a market's theme and then pull its own event coverage with list_rwa_news `catalyst`. Each market also carries `venueClosedAt`, VENUE-SOURCED only: non-null when the venue has reported the market closed but the winning outcome was not yet readable, meaning the market is "settling" — awaiting resolution, so its `prob`/`delta24h` are frozen at the last observed value rather than a live quote. Never inferred from `asOf` age; a market that is merely old but still open on its venue always reports null here. This is observed venue pricing, not a forecast. If the response carries `smartFlowDegraded: true`, the smart-money flow inputs could not be read on that compute: the absence of `smartFlow` on these rows then says nothing about the last 24h, so report that the flow data is temporarily unavailable rather than that no wallets have been buying. Even when it is false, `smartFlow` is computed only for the top-50 Polymarket markets by volume, so its absence on a Kalshi row or a lower-volume Polymarket row means NOT EVALUATED — never "no wallet bought it". Start here to find the markets a question is about, then use get_market_context or get_venue_calibration to understand what a quoted probability is worth.

Backing endpoint: /api/markets

Get Market Range Position get_market_context

Returns where one market's current YES probability sits within its own trailing 30-day distribution: percentile, z-score, and the window low / median / high. This describes the market's recent range — it is NOT a claim that the market is mispriced. A market near the top of its range has moved, which is a fact about its history, not a signal about its future. A window with fewer than 30 observations returns insufficientContext rather than a derived number — but check `degraded` before attributing it to a thin history: when that is true the observation series could not be READ at all, so report the context as temporarily unavailable rather than saying Argos has few observations for the market. Two honesty flags qualify the reading: `dislocated` (the current value has broken OUTSIDE its historical range — the percentile/trend then describe where it was, and the trend reports the breakout direction) and `flatDistribution` (the range is negligible, so a percentile near 0 or 1 is not meaningful unless the value has dislocated). Requires a marketId + venue from list_prediction_markets.

Backing endpoint: /api/market-context

List the Daily Prediction-Market Brief list_predict_brief

Returns the ranked daily brief across the tracked market universe: the markets with the widest range moves, the markets resolving soonest, and the largest 24h volume. A summary of where the universe moved, expressed as observations about market movement rather than recommendations. Each range/move/resolving card carries per-card triage fields: volUsd/vol24h with volUsdSemantics/quoteQuality (same source + meaning as list_prediction_markets), plus asOf + coverage from the context envelope (hourly-rollup granularity, so a card's asOf can lag the raw-observation asOf on list_prediction_markets by ~1h — both are honest observation times). Triage a card for freshness and liquidity without a second call. If `meta.degraded` is true the market history could not be READ on that compute: `meta.ranked` goes to 0, the scores report an unavailable verdict (or are absent entirely), AND every empty card's `reason` says the same. Report the brief as temporarily unavailable rather than saying no market has enough history — that phrasing is a claim about Argos's corpus, and it would be wrong. When `meta.degraded` is absent or false, ranked/insufficient and the card reasons are a real reading. Use when the user asks "what is moving in prediction markets right now" and has not named a specific market.

Backing endpoint: /api/predict-brief

No arguments required.

List Markets Where Profitable-History Wallets Hold a Position list_smart_money_plays

Returns markets where a cohort of wallets with a profitable trading history CURRENTLY HOLDS a position — flavor "carry" (holding a near-certain favorite or longshot to resolution) or "value" (a directional position away from the price extremes), with the held side. (A few rows are cohort-less price-dislocation signals instead — flavor "sharp_value" vs a sharp-book consensus, or "cross_venue" a gap to another venue — these carry no cohort.) The cohort rows are a snapshot of HELD positions, not recent trading — do not conflate it with the `smartFlow` field on list_prediction_markets, which is the recent 24h net-BUY flow direction. The two can legitimately differ (a cohort can hold one side while the last 24h of flow leans the other, e.g. profit-taking); that is not a contradiction, they measure holdings vs recent flow. It reports what a cohort holds, not what the user should do. A wallet's past profitability does not predict its future results, and Argos research found that naively copying high-return wallets did not beat market consensus out of sample. Treat this as one observation among several, never as a recommendation. If the response carries `degraded: true`, the upstream wallet data could not be read on that compute: an empty `rows` then says nothing about the markets, so report that the data is temporarily unavailable rather than that no positions are held.

Backing endpoint: /api/predict-copyable

No arguments required.

List Cross-Venue Gaps and Basket Sums list_cross_venue_arbitrage

Returns two mechanical pricing observations. `rows`: price gaps for the same event quoted on two different venues, net of fees. `baskets`: mutually-exclusive market sets — for example the "which team wins?" legs of one tournament — whose tracked YES probabilities sum ABOVE 100% (an overround on the observed legs). Sums below 100% are never flagged: the tracked legs can be a subset of the full outcome set, so a low sum may reflect coverage rather than mispricing. A basket sum is an arithmetic identity across the tracked legs, not a prediction about any single leg, and a set with any unpriced leg is returned probability-only. Gaps are computed from quoted prices and do not account for order-book depth, slippage, or the capital cost of holding to resolution, so a quoted gap is not a realizable profit. The response carries two INDEPENDENT degraded flags: if `arbDegraded` is true the equivalence read failed, so an empty `rows` says nothing about how many pairs are confirmed; if `basketsDegraded` is true the basket read failed, so an empty `baskets` says nothing about current dislocations. In either case report that half as temporarily unavailable rather than as an absence — and never let one flag qualify the other's field.

Backing endpoint: /api/predict-arb

No arguments required.

Get How Well Each Venue Has Been Calibrated get_venue_calibration

Returns each venue's Brier score and bucketed outcome rates over the market resolutions Argos has captured — i.e. of the markets a venue priced near 70%, what share actually resolved YES. Call this before characterizing any quoted probability: it grounds the quote in how that venue's probabilities have historically resolved, instead of treating the quote as truth. Each venue reports `n` (resolutions captured) and `sufficient` against the top-level `minSampleN` threshold: when `sufficient` is false the sample is too thin to rely on — do NOT convert a quote into a calibrated belief off that venue, even though the real (thin) Brier is still shown. Returns an empty venue list when no resolutions have been captured, rather than a fabricated score — but check `degraded` first: when it is true the corpus could not be read on that compute, so an empty list says nothing about how many resolutions exist. Report the calibration data as temporarily unavailable rather than as an absence, and do not treat the quote as ungroundable on that basis alone. `degraded` is a DIFFERENT statement from a returned venue's `sufficient:false`: that one is a real reading of a thin sample, this one means there was no reading at all.

Backing endpoint: /api/predict-calibration

No arguments required.