Changelog

What shipped —
since day one.

Every significant build milestone, from the first public commit (Feb 8, 2026) to today. Data-first, no hype. All accuracy claims auditable at /v1/stats.

July 2026
Agent payments, provider redundancy & attribution close-out
  • Agent-native payments live on mainnet. /x402/v1/* — autonomous agents pay per-call in USDC, no signup, no key. Curated endpoints exposed via the x402 standard; submitted to Solana's pay-skills registry.
  • Provider redundancy + failover. Multi-provider RPC routing with automatic failover and a metered budget guard across providers — resilience and lower latency on hot-wallet reads.
  • Public Telegram channel + alert bot. t.me/solsentryai — CRITICAL alerts and an on-demand /scan command; invite-based DM onboarding.
  • Attribution correction complete. The multi-source re-verification of historical dev-wallet attributions is finished and deployed; inflated legacy counts corrected. Honest data over big numbers.
June 2026
Hardening, low-latency layer & launch prep
  • Public site live. The SolSentry site went officially online — about, changelog, public API reference, and a live operator lookup, with every accuracy claim linking straight to /v1/stats.
  • Attribution audit complete. Full multi-source re-verification of historical dev-wallet attributions, cross-checked across independent on-chain datasets. Honest data over big numbers.
  • Classifier self-correction. Found and fixed a false-positive class (position-NFTs read as tokens); backfilled affected predictions. CRITICAL-tier precision now 88.3%, auditable per-mint at /v1/predictions/{mint}.
  • Shadow detectors wired. Forensic detectors (cashout tracing, CEX-attribution, sybil-evasion patterns) now run in shadow mode — observing, not yet scoring — to validate signal before going live.
  • Cost-control & reliability. Self-detecting waste guardrail (a misconfigured webhook burned credits overnight → velocity gate + event-type filter + daily reconcile). Self-healer fixed to run on schedule. Redis-backed rate limiting + Postgres auth layer landed ahead of launch.
  • /v1/contract-analysis — program/contract analysis via the API.
May 2026
Case machine, dataset-first architecture & product surface
  • Dataset-first architecture. Unified waterfall resolver (dataset-first → RPC fallback) + RPC-level cache, cutting repeated on-chain spend on re-queried mints.
  • Bidirectional tracer + SNS deep-trace. Full Solana Name Service ownership graph; multi-hop fund-flow investigator following outflow to its terminus.
  • /v1/external-history — unified endpoint aggregating transaction history across providers.
  • Adversarial monitor + narrative generator. Autonomous prose from investigation data; 7 alert types; structured-deposit ("smurfing") cluster detection.
  • @solsentry/guard SDK v0 — pre-signing risk client published to NPM (MIT, examples, CI template).
  • Next.js 15 public site MVP + documentation scaffold.
  • Submitted to Colosseum Frontier (May 12).
April 2026
Operator graph, ALife brain & public API
  • Public API launch + @solsentry/mcp on NPM. Free REST API and an MCP server exposing operator risk as a composable primitive. Open-core surface live (MIT SDK, SECURITY policy, public roadmap).
  • ALife autonomous brain ships. Evolutionary agents on live mainnet: signal feedback loop, retraction engine, anomaly seeker, and a self-healer for autonomous data-integrity repair.
  • Operator graph + investigation graph. Operator → token → wallet relationships connected; /v1/graph (named nodes, roles), /v1/drain-trace/{wallet} fund-flow tracing, ring detector.
  • Token-2022 support + launchpad detection. Native Token-2022 handling; 12 launch platforms recognized; King-of-the-Hill pattern detection.
  • Multi-window outcome resolution. 24h / 3d / 7d resolution windows + fast-track; accuracy feedback wired end-to-end. Live accuracy auditable at /v1/stats.
March 2026
Accuracy feedback loop & operator profiling
  • Outcome tracker wired — accuracy feedback loop active. Resolved predictions feed back into the classifier; first build with a live, self-correcting accuracy metric.
  • Hunter auto-scan pipeline. Automated high-risk detection routed without manual trigger.
  • Operator profiling + KOL cross-reference. Per-wallet context and holder-confidence tiers in scan output; evolutionary feedback persists across restarts.
February 2026
First scanner, agents & on-chain monitoring
  • First public commit (Feb 8) — token scanner, outcome tracker, baseline intelligence pipeline.
  • Autonomous Hunter & Sentinel agents. Watch for new deploys, large transfers, and liquidity removals; first evolutionary genome module.
  • RPC pool with failover + holder-distribution analysis.
  • Known-entities + wallet-alias registry. Seeded registry of known programs/protocols to suppress false positives and add context; operator-to-alias mapping.

Capability anchors

Free REST API ↗
No key, no signup. /v1/operator/{wallet}: risk level, confirmed rugs, serial-deployer flag, connected bot clusters. Sub-second scan response.
MCP server ↗
@solsentry/mcp on NPM. Operator risk as a composable primitive.
Auditable ↗
/v1/predictions/{mint} and /v1/operator/{wallet} are public. Check any claim.
Open-core ↗
Open SDK + MCP + free API; the intelligence engine is proprietary.