Data & Public Claims
Conyr publishes the numbers an agent may rely on with their measurement date, counting rule, scope, evidence, and correct interpretation. Implementation counts are kept separate from dated production measurements so an agent never treats an audited snapshot as a live counter.
Download the machine-readable registry → · Read the coverage contract → · Open the benchmark →
The registry is public-safe by design. It includes enough methodology to interpret and cite each aggregate while excluding credentials, internal hosts, private query shapes, and database infrastructure.
Reproducible product facts
These values are checked against code, configuration, or the public product contract during the frontend build. A changed catalog or registry breaks the check until the public claim is reviewed.
Implementation verifiedIngestion · as of Aug 19, 2026
25
25 Solana trading programs in the parser registry
- Scope
- Unique on-chain trading program IDs in the indexer's active DEX program registry.
- How it is checked
- Count unique entries in the active parser registry. The build-time claim check compares this public value with the implementation.
- Observation window
- Current implementation at the as-of date.
- Interpretation
- Program IDs are an implementation count; protocol families can include multiple deployments.
- Use endpoint response semantics—not registry membership alone—to determine whether a requested event is available.
Citation-ready: Conyr's parser registry contained 25 Solana trading program IDs on August 19, 2026.
Implementation verifiedDelivery · as of Aug 19, 2026
46
46 tools exposed by the hosted Conyr MCP server
- Scope
- External tools returned by the MCP catalog: 42 focused tools and four composite dossier tools. Internal UI-only hooks are excluded.
- How it is checked
- Count the externally exposed focused and composite tool specifications after applying the MCP exclusion list. The build-time claim check compares this value with the implementation.
- Observation window
- Current implementation at the as-of date.
- Interpretation
- Tool count describes catalog breadth, not independent underlying datasets.
- Access to individual tools depends on the caller's API tier.
Citation-ready: Conyr exposed 46 hosted MCP tools—42 focused tools and four composite dossiers—on August 19, 2026.
Implementation verifiedDelivery · as of Aug 25, 2026
41
41 documented REST operations under the versioned API
- Scope
- Method-and-path operations documented beneath the versioned /v1 REST surface; system and web-app routes are excluded.
- How it is checked
- Count documented HTTP operations beneath /v1 in the public OpenAPI contract. The build-time claim check compares this value with that contract.
- Observation window
- Current public OpenAPI contract at the as-of date.
- Interpretation
- Operation count describes interface breadth, not the number of independent datasets.
- Tier gates, rate limits, parameters, and null semantics still apply per operation.
Citation-ready: Conyr's public OpenAPI contract documented 41 versioned REST operations on August 25, 2026.
Implementation verifiedDelivery · as of Aug 19, 2026
13
13 WebSocket channel families
- Scope
- Distinct client-facing subscription patterns for token and wallet event families; parameter variants such as OHLCV timeframe do not create extra families.
- How it is checked
- Count accepted token and wallet channel patterns in the WebSocket protocol mapper. The build-time claim check compares this value with the implementation.
- Observation window
- Current client-facing protocol at the as-of date.
- Interpretation
- WebSocket delivery favors low latency and is not a replayable guaranteed-delivery log.
- Channel access and connection limits depend on the caller's API tier.
Citation-ready: Conyr's client-facing WebSocket protocol exposed 13 channel families on August 19, 2026.
Implementation verifiedFunding provenance · as of Aug 26, 2026
128
128 curated funding-origin addresses
- Scope
- Unique addresses in the maintained provenance-root registry, spanning exchange, custody, payment, service, and other recognized infrastructure origins.
- How it is checked
- Count unique address keys in the curated provenance-root registry. The build-time claim check compares this value with the maintained configuration.
- Observation window
- Current curated registry at the as-of date.
- Interpretation
- The registry intentionally includes recognized infrastructure that is not a centralized exchange.
- Labels require curation and can lag address rotation or newly identified service wallets.
Citation-ready: Conyr's curated provenance registry contained 128 recognized funding-origin addresses on August 26, 2026.
Implementation verifiedFunding provenance · as of Aug 19, 2026
20 hops
Funding provenance traced up to 20 hops
- Scope
- Maximum backward funding-path depth exposed by the wallet-provenance product.
- How it is checked
- Verify the configured product limit and public API contract. The claim describes a maximum search depth, not guaranteed resolution depth.
- Observation window
- Current product limit at the as-of date.
- Interpretation
- A path can terminate earlier at a known origin, missing transfer, ambiguity, or unavailable history.
- No match within 20 hops means unresolved, not that the wallet has no funder.
Citation-ready: Conyr's wallet-provenance product searches backward through as many as 20 funding hops.
Implementation verifiedWallet intelligence · as of Aug 19, 2026
30-dim
30 dimensions in each eligible wallet behavior fingerprint
- Scope
- Named features in the wallet behavior vector used by the similarity service.
- How it is checked
- Count the canonical feature-name array and compare it with the configured vector dimension during the build-time claim check.
- Observation window
- Current implementation at the as-of date.
- Interpretation
- Only activity-gated wallets receive a vector.
- Dimensions are engineered behavioral features, not independent causal factors.
Citation-ready: Conyr represented eligible wallet behavior with a 30-dimensional feature vector on August 19, 2026.
Implementation verifiedToken intelligence · as of Aug 19, 2026
6
6 confidence-weighted token health dimensions
- Scope
- Independently recomputed token-health dimensions: security, liquidity, distribution, momentum, trader quality, and behavior.
- How it is checked
- Count the canonical token-score dimension variants. The build-time claim check compares this value with the implementation.
- Observation window
- Current implementation at the as-of date.
- Interpretation
- The composite is confidence-weighted; a missing dimension does not contribute a neutral score.
- The score summarizes observed evidence and is not a guarantee, accusation, or trading recommendation.
Citation-ready: Conyr's token health model used 6 confidence-weighted dimensions on August 19, 2026.
Implementation verifiedCoordination · as of Aug 19, 2026
10 classes
10 structural bundle classifications
- Scope
- Structural variants represented by the bundle type used by detection and lifecycle tracking.
- How it is checked
- Count the variants in the canonical bundle classification type. The build-time claim check compares this value with the implementation.
- Observation window
- Current implementation at the as-of date.
- Interpretation
- The ten structural classes are consolidated into five headline risk archetypes on some product surfaces.
- Classification is evidence-based detection, not proof of common beneficial ownership or malicious intent.
Citation-ready: Conyr's bundle engine represented 10 structural classifications on August 19, 2026.
Published production measurements
These are dated aggregate measurements from Conyr's production corpus. Each retains its scope, method, observation window, and interpretation rules; it is evidence of demonstrated scale, not a live counter.
Published measurementIngestion · as of June 2026
20M+
More than 20 million Solana swaps parsed per day
- Scope
- Successfully decoded Solana swap events processed by Conyr's first-party transaction pipeline; failed transactions and non-swap activity are outside this count.
- How it is checked
- Count decoded swap events in the production observation window, normalize the covered interval to a daily rate, and publish a rounded-down lower bound.
- Observation window
- Daily rate observed within a trailing approximately 20-day production window.
- Interpretation
- This is a dated production measurement, not a promise that every future day exceeds the threshold.
- The count is decoded swaps, not all Solana transactions or unique trades after wallet-level deduplication.
Citation-ready: Conyr measured more than 20 million decoded Solana swaps per day in a June 2026 production window.
Published measurementCorpus coverage · as of June 2026
3.4M+
More than 3.4 million active wallets in the measured corpus
- Scope
- Unique wallet addresses observed in qualifying decoded trading activity during the window.
- How it is checked
- Deduplicate qualifying trader addresses over the observation window and publish a rounded-down lower bound.
- Observation window
- Trailing approximately 20-day production window.
- Interpretation
- A wallet is an address, not necessarily one human or legal entity.
- Active-wallet counts are window-dependent and should not be read as all-time users.
- Evidence
- Wallet coverage definition ↗
Citation-ready: Conyr observed more than 3.4 million active wallet addresses in its June 2026 production window.
Published measurementCorpus coverage · as of June 2026
660K+
More than 660,000 traded token mints in the measured corpus
- Scope
- Distinct Solana token mints present in qualifying decoded swap activity during the window.
- How it is checked
- Deduplicate token mints represented in decoded swaps and publish a rounded-down lower bound.
- Observation window
- Trailing approximately 20-day production window.
- Interpretation
- Inclusion means observed trading activity, not endorsement, liquidity, safety, or complete metadata.
- Counts vary with the observation window and token creation activity.
- Evidence
- Token coverage definition ↗
Citation-ready: Conyr observed more than 660,000 traded Solana token mints in its June 2026 production window.
Published measurementFunding provenance · as of June 2026
250M+
More than 250 million funding-provenance transfers resolved
- Scope
- Funding-transfer relationships available to the provenance pipeline, including paths that do not terminate at a recognized origin.
- How it is checked
- Count normalized funding-transfer relationships available during the observation window and publish a rounded-down lower bound.
- Observation window
- Trailing approximately 20-day production window.
- Interpretation
- A resolved transfer relationship is not the same as a wallet being attributed to a known entity.
- Unknown, ambiguous, or depth-capped paths remain explicitly unresolved.
Citation-ready: Conyr's June 2026 production snapshot contained more than 250 million funding-provenance transfer relationships.
Published measurementCoordination · as of June 2026
2.7M+
More than 2.7 million leader-to-follower edges
- Scope
- Directed wallet relationships retained by the coordination engine after timing, shared-asset, support, and trade-size evidence are scored.
- How it is checked
- Count retained directed relationships in the measured coordination snapshot and publish a rounded-down lower bound.
- Observation window
- Trailing approximately 20-day production window.
- Interpretation
- An edge is evidence of repeat temporal following, not proof of ownership, intent, or off-chain coordination.
- Counts depend on retention, scoring thresholds, and the measured window.
Citation-ready: Conyr retained more than 2.7 million scored leader-to-follower edges in its June 2026 coordination snapshot.
Published measurementCoordination · as of June 2026
160K+
More than 160,000 time-bounded token campaigns detected
- Scope
- Time-bounded token-level coordination episodes emitted by the campaign detector.
- How it is checked
- Count distinct campaign records in the measured coordination snapshot and publish a rounded-down lower bound.
- Observation window
- Trailing approximately 20-day production window.
- Interpretation
- Campaigns describe statistically unusual synchronized activity; they are not accusations or trading signals.
- One token can have multiple campaigns in separate time windows.
Citation-ready: Conyr detected more than 160,000 time-bounded token campaigns in its June 2026 production snapshot.
Published measurementWallet intelligence · as of June 2026
950K+
More than 950,000 wallets with computed label sets
- Scope
- Wallets meeting at least one classification rule in the measured label snapshot.
- How it is checked
- Count distinct wallets carrying computed behavioral or provenance labels and publish a rounded-down lower bound.
- Observation window
- Trailing approximately 20-day production window.
- Interpretation
- Labels are computed from observed behavior and can change as new activity arrives.
- A label is a classification signal, not identity proof or a guarantee of future performance.
Citation-ready: Conyr computed label sets for more than 950,000 wallets in its June 2026 production snapshot.
Published measurementWallet intelligence · as of June 2026
33M+
More than 33 million PnL accounting events
- Scope
- Position-accounting events produced from qualifying decoded wallet trades after accounting exclusions.
- How it is checked
- Count emitted PnL accounting events in the measured snapshot after flash-swap and accounting filters, then publish a rounded-down lower bound.
- Observation window
- Trailing approximately 20-day production window.
- Interpretation
- Event count is not the number of closed positions, wallets, or profitable trades.
- Transfer-only inventory and unsupported trade paths can produce partial or unavailable cost basis.
Citation-ready: Conyr's June 2026 production snapshot contained more than 33 million position-accounting PnL events.
Published measurementWallet intelligence · as of June 2026
860K+
More than 860,000 wallets represented in PnL accounting
- Scope
- Distinct wallet addresses with qualifying position-accounting data in the measured snapshot.
- How it is checked
- Deduplicate wallet addresses represented by qualifying PnL records and publish a rounded-down lower bound.
- Observation window
- Trailing approximately 20-day production window.
- Interpretation
- Coverage does not imply every token position for each wallet has complete cost basis.
- Wallets with too little qualifying activity may lack higher-level behavioral profiles.
Citation-ready: Conyr represented more than 860,000 wallets in its June 2026 PnL snapshot.
Published measurementWallet intelligence · as of June 2026
200K+
More than 200,000 activity-gated wallet behavior fingerprints
- Scope
- Wallets that passed the minimum closed-position, activity-span, and trade-count gates and had a behavior vector in the measured index.
- How it is checked
- Count vectors in the production similarity index after activity gates and publish a rounded-down lower bound.
- Observation window
- Production vector index snapshot.
- Interpretation
- The activity gate intentionally excludes thin-history wallets rather than assigning them low-confidence vectors.
- Similarity means behavioral resemblance in the modeled dimensions, not common ownership or coordination.
Citation-ready: Conyr indexed more than 200,000 activity-gated wallet behavior fingerprints in its June 2026 snapshot.
Published measurementCoordination · as of June 2026
600K+
More than 600,000 swaps linked to detected bundle structures
- Scope
- Decoded swaps attached to a retained coordinated-wallet structure by the bundle detector.
- How it is checked
- Count swap records associated with retained bundle structures in the measured window and publish a rounded-down lower bound.
- Observation window
- Trailing production window reported in June 2026.
- Interpretation
- A bundle-linked swap is a detection result, not proof that a named person controlled the wallets.
- Detector thresholds and supported execution patterns bound recall.
Citation-ready: Conyr linked more than 600,000 swaps to detected bundle structures in a June 2026 trailing production window.
Published measurementWallet intelligence · as of Aug 19, 2026
194,610
194,610-wallet temporal study of closed-lot win rate
- Scope
- Wallets with at least 10 decided fully closed FIFO entry lots in formation and at least 3 lots both entered and fully closed in evaluation after explicit bot, arbitrage, malicious-wallet, and duplicate-event exclusions.
- How it is checked
- Rank eligible wallets by formation closed-lot win rate, compare the top and bottom deciles on positive evaluation-period cumulative gross realized PnL from new evaluation-window lots, and publish counts, cutoffs, medians, 95% Wilson intervals, an exit-row sensitivity result, and closed-only limitations.
- Observation window
- A 13-day formation window from July 24 to August 6, 2026, followed by a 13-day evaluation window through August 19, 2026.
- Interpretation
- The top-decile closed-lot cohort had a 60.70% positive subsequent cumulative-gross-realized-PnL rate versus 13.44% for the bottom decile, but this is an observational association rather than a causal effect.
- The exit-row sensitivity result was similarly strong and slightly wider in this window; the study does not claim closed-lot win rate was the superior predictor.
- Open, abandoned, transferred-out, and never-sold inventory was not marked to market, creating material closed-only survivorship bias.
- The cumulative production lot field is gross of separately allocated entry and exit network-fee fields, so the outcome is not fee-net profitability.
- Eligibility conditions on evaluation-period activity, entry, and closure, so the result does not describe all wallets observed in formation.
Citation-ready: In Conyr's July 24–August 19, 2026 temporal study of 194,610 eligible Solana wallets, the top formation closed-lot-win-rate decile had positive subsequent cumulative gross realized PnL at 60.70%, versus 13.44% for the bottom decile; the closed-only, fee-gross, and observational limitations apply.
Citation contract
Cite Conyr, the claim title, its as-of date, and the canonical claim anchor. Preserve the claim’s unit and scope: a wallet address is not necessarily one person, detected coordination is not a declaration of intent, and a dated production measurement is not a live counter.
Conyr. “46 tools exposed by the hosted Conyr MCP server.”
As of 2026-08-19. https://conyr.ai/docs/agents/data#claim-mcp-tools