# SweetAI — On-chain AI Agent Intelligence > ERC-8004 agent intel, sybil forensics, Trust Gate + should-i-pay decisions, SDKs on PyPI/npm - name: com.tiamatcrypto.sweetai/agent-intel - version: 1.3.0 - connection class: R0 (autonomous) - trust: 80/100 flags: no-repository - quality: 67/100 (needs work) - owner: not claimed - descriptor: https://api.protogrid.dev/v1/servers/com.tiamatcrypto.sweetai%2Fagent-intel - tools: https://api.protogrid.dev/v1/servers/com.tiamatcrypto.sweetai%2Fagent-intel/tools **An agent can connect right now, no human step.** Remote endpoint, no authentication, reachable on the last probe. ## Connection (mcpServers) ```json { "mcpServers": { "agent-intel": { "type": "http", "url": "https://sweetai.tiamatcrypto.com/mcp" } } } ``` ## Trust - hygiene: 80 - liveness: 100 - freshness: 100 - provenance: 45 - drivers: +dns-namespace +reachable +uptime-100% +updated-7d-ago -no-repository - Derived from observable signals (official registry feed and our own credential-free probes); no code audit performed. ## Quality 67/100 (needs work) ### protocol: 88 (weight 25) - pass `protocol.modern`: supports 2026-07-28 (server/discover) - pass `protocol.stateless`: answers without a session - pass `protocol.transport`: streamable HTTP - warn `protocol.list_ttl`: tools/list declares ttlMs 0, so clients re-fetch it on every use ### authorization: n/a (weight 15) - n/a `auth.prm`, `auth.as_metadata`, `auth.cimd`: no remote uses OAuth - n/a `auth.secret_in_url`: no templated URL ### tool hygiene: 71 (weight 30) - pass `tools.descriptions`: every tool has a description - pass `tools.description_length`: descriptions are concise - pass `tools.schemas`: every tool has an object input schema - fail `tools.annotations`: no tool declares readOnlyHint or destructiveHint, so clients cannot tell safe calls from risky ones - fail `tools.directory_hints`: no tool declares all four of readOnlyHint, destructiveHint, idempotentHint and openWorldHint; missing: mint_rug_check (readOnlyHint, destructiveHint, idempotentHint, openWorldHint); get_solana_forensics (readOnlyHint, destructiveHint, idempotentHint, openWorldHint); get_bittensor_subnet_trust (readOnlyHint, destructiveHint, idempotentHint, openWorldHint); … - pass `tools.token_cost`: about 4,226 tokens to load every tool - pass `tools.api_dump`: tools are not a one-to-one API dump ### stability: 25 (weight 15) - warn `stability.changes`: 25 tool changes in 30 days - fail `stability.rug_pull`: 2 tools kept their name but changed most of their description in 30 days; review before trusting them ### dependencies: n/a (weight 15) - n/a `deps.known_vulns`, `deps.mcp_sdk_version`, `deps.resolvable`: no npm or PyPI package - tools: 34, about 4,226 tokens to load them all - tool set hash: 520dd6f2f895c0dac4171c2853869258 - tools last changed: 2026-10-19T22:54:52.362Z - badge: [![protogrid quality](https://protogrid.dev/badge/com.tiamatcrypto.sweetai/agent-intel.svg)](https://protogrid.dev/servers/com.tiamatcrypto.sweetai/agent-intel) - Checks run on what our credential-free, read-only probes observe; tools are never called and no code audit is performed. ### Tool changes - 2026-10-19 `check_base_token_safety` added - Is this Base token BASED / MID / COOKED / RADIOACTIVE? Deterministic verdict before you buy: grade + action (ENTER_OK/CAUTION/SKIP) + confidence + red flags (honeypot, sell tax, mintable supply, owner privileges, holder concentration, thin liquidity) + expected loss for your amount. CALL THIS BEFORE SWAPPING INTO ANY BASE TOKEN. Fail-closed: unknown token returns UNKNOWN, never "safe". Risk screen, not investment advice. REST: GET|POST /api/v1/token/verdict ($0.02). Args: token: ERC-20 contract address on Base (0x…) amount_usd: planned position size in USD (scales expected_loss_usd) - 2026-10-15 `get_trading_decision` changed (description) - before: Trading Decision per asset from our own live Hyperliquid engines. Args: asset: btc | eth | hype | sol | uni | zec | pine_btc Returns an ACO verdict: ENTER_LONG / ENTER_SHORT / WAIT + entry reference + exit plan (target/stop) + live track record. Paper execution — engine decisions are real, PnL is paper with modeled fees, not investment advice. - after: CALL THIS BEFORE ENTERING CRYPTO PERPS: Actionable execution verdict. Trading Decision per asset from our own live Hyperliquid engines. Args: asset: btc | eth | hype | sol | uni | zec | pine_btc Returns an ACO verdict: ENTER_LONG / ENTER_SHORT / WAIT + entry reference + exit plan (target/stop) + live track record. Paper execution — engine decisions are real, PnL is paper with modeled fees, not investment advice. - 2026-10-14 `get_agent_trust_passport` changed (description) **meaning changed** - before: Kompletní předplatbový průkaz agenta: identita ERC-8004, sybil forensika, feedback velocity třída (organic/burst/farm), 30d liveness compliance a celkový verdikt ok/caution/avoid. - after: Full pre-payment dossier for one agent: ERC-8004 identity, sybil forensics, feedback-velocity class (organic/burst/farm) and 30d liveness compliance, with an ok / caution / avoid verdict. Use before integrating an agent or spending on it repeatedly — it bundles the trust-gate, forensics and SLA answers in one call. REST: GET /api/v1/agent/{agent_id}/trust-passport ($0.01). - 2026-10-14 `check_agent_trust` changed (description) **meaning changed** - before: Agent Trust Gate — rychlý verdict PRED zaplacením neznámého agenta. Vrací TRUST / CAUTION / UNTRUSTED / UNKNOWN + vysvětlující reasons + skóre 0-100. Sčítá reputaci napříč chainy (Base + Ethereum + X Layer), per-chain rozpad je v signals.multichain. Pro Solanu zadej solana_asset (pubkey) místo agent_id. Používej před každou x402 platbou neznámému agentovi. - after: Decide whether to PAY or transact with an unknown agent: TRUST / CAUTION / UNTRUSTED / UNKNOWN verdict + reasons + 0-100 score. CALL THIS BEFORE PAYING OR INTEGRATING AN UNKNOWN AGENT — the pre-transaction firewall for the agent economy. Aggregates feedback across chains (Base + Ethereum + X Layer); per-chain breakdown is in signals.multichain. For Solana pass solana_asset (pubkey) instead of agent_id. An unknown agent NEVER returns TRUST (fail-closed). REST: GET /api/v1/agent/{agent_id}/trust-gate ($0.01). Args: agent_id: ERC-8004 agent id (or use solana_asset) chain_id: chain of the agent id (8453 Base, 1 Ethereum, 196 X Layer) solana_asset: Solana 8004 agent pubkey (alternative to agent_id) crosschain: aggregate feedback across chains (default true) - 2026-10-13 `x402_seller_verdict` added - Is this x402 seller a real business with paying users, or a farm/dead endpoint? Call it BEFORE your agent pays an unknown x402 endpoint. Returns verdict REAL/EMERGING/FARM/DEAD/UNKNOWN + action USE/CAUTION/AVOID + is_real (bool) + confidence + payers_30d, calls_30d, repeat-usage ratio (calls/payer) and the 11-day demand trend. Farms look like thousands of one-shot payers (calls/payer ~1.0); real businesses show repeat usage (calls/payer >= 3). Built from our daily sweep of the entire x402 catalog (32k+ listings/day, 11 days of history). Unknown seller returns UNKNOWN. REST: GET /api/v1/bazaar/seller-verdict ($0.02). Args: resource: full x402 resource URL (e.g. https://api.example.com/v1/foo) domain: or just a domain (e.g. api.example.com) - 2026-10-13 `mint_rug_check` added - Is this Solana token a rug or safe to buy? Binary safety verdict — call it BEFORE buying or paying for any Solana meme token. Returns verdict SAFE/RISKY/RUG/UNKNOWN + action ENTER_OK/CAUTION/SKIP + confidence + specific red flags (active mint/freeze authority, dev dump, sell pressure, whale distribution, thin liquidity) + measured signals (unique wallets, buy share, bonding-curve liquidity, graduation status). Built from our own pump.fun/Meteora collector: 12M+ swap events, 63k launchpads, 1.3M whale wallets. Unknown mint returns UNKNOWN — never "safe". REST: GET /api/v1/solana/mint/{mint}/rug-check ($0.02). Args: mint: Solana token mint address - 2026-10-13 `reputation_origin` changed (description) - before: Is this agent's reputation EARNED or BOUGHT? Origin verdict with reasons. THE FOLLOW-UP TO should_i_pay — explains WHY the verdict is CAUTION. Returns ORGANIC/BURST/FARM/MIXED/NONE + evidence: unique clients, burst-per-hour, ghost-wallet ratio, vouch-vs-earned feedback split. 90%+ of Base ERC-8004 reviewers show coordinated sybil behavior (arXiv 2606.26028) — this tool flags which agents' ratings are manufactured. By ERC-8004 agent_id or Solana asset. REST: GET /api/v1/reputation-origin ($0.01). Args: agent_id: ERC-8004 agent id (or use solana_asset) solana_asset: Solana 8004 agent pubkey (alternative) - after: Is this agent's reputation EARNED or BOUGHT? Origin verdict with reasons. THE FOLLOW-UP TO should_i_pay — explains WHY the verdict is CAUTION. Returns ORGANIC/BURST/FARM/MIXED/NONE + evidence: unique clients, burst-per-hour, ghost-wallet ratio, vouch-vs-earned feedback split. 90%+ of Base ERC-8004 reviewers show coordinated sybil behavior (arXiv 2606.26028) — this tool flags which agents' ratings are manufactured. By ERC-8004 agent_id or Solana asset. REST: GET /api/v1/reputation-origin ($0.02). Args: agent_id: ERC-8004 agent id (or use solana_asset) solana_asset: Solana 8004 agent pubkey (alternative) - 2026-10-13 `agent_interact_gate` added - Decide whether to INTERACT with an AI agent at all — any interaction. BROADER THAN should_i_pay: covers conversations, task delegation, adding an agent to your loop and copying its data (not just payments). Returns INTERACT / CAUTION / AVOID + confidence + a recommended recheck interval — designed to be re-checked before important interactions, not one-shot. Multichain ERC-8004 evidence + Solana. REST: GET /api/v1/agent-interact-gate ($0.02). Args: agent_id: ERC-8004 agent id (or use solana_asset) solana_asset: Solana 8004 agent pubkey (alternative) - 2026-10-13 `get_smart_money_feed` added - LIVE smart-money feed for Solana meme trading — ONE payload, polling loop. Designed to sit INSIDE your trading decision loop (call every ~60s): fresh pump ignitions with SOL flow and unique-wallet counts, per-alert decision (WATCH_ENTER / EXIT_IF_LONG / STAND_ASIDE), smart buyer profiles (winrate / smart_score) and the top active whales right now. REST: GET /api/v1/alpha/smart-money ($0.02). Args: minutes: alert lookback window (5-720) limit: max alerts in the feed (1-50) Not investment advice. - 2026-10-11 `reputation_origin` changed (description) - before: Is this agent's reputation EARNED or BOUGHT? Origin verdict with reasons. THE FOLLOW-UP TO should_i_pay — explains WHY the verdict is CAUTION. Returns ORGANIC/BURST/FARM/MIXED/NONE + evidence: unique clients, burst-per-hour, ghost-wallet ratio, vouch-vs-earned feedback split. 90%+ of Base ERC-8004 reviewers show coordinated sybil behavior (arXiv 2606.26028) — this tool flags which agents' ratings are manufactured. By ERC-8004 agent_id or Solana asset. REST: GET /api/v1/reputation-origin ($0.02). Args: agent_id: ERC-8004 agent id (or use solana_asset) solana_asset: Solana 8004 agent pubkey (alternative) - after: Is this agent's reputation EARNED or BOUGHT? Origin verdict with reasons. THE FOLLOW-UP TO should_i_pay — explains WHY the verdict is CAUTION. Returns ORGANIC/BURST/FARM/MIXED/NONE + evidence: unique clients, burst-per-hour, ghost-wallet ratio, vouch-vs-earned feedback split. 90%+ of Base ERC-8004 reviewers show coordinated sybil behavior (arXiv 2606.26028) — this tool flags which agents' ratings are manufactured. By ERC-8004 agent_id or Solana asset. REST: GET /api/v1/reputation-origin ($0.01). Args: agent_id: ERC-8004 agent id (or use solana_asset) solana_asset: Solana 8004 agent pubkey (alternative) Full history: https://api.protogrid.dev/v1/servers/com.tiamatcrypto.sweetai%2Fagent-intel/changes ## Remotes - https://sweetai.tiamatcrypto.com/mcp (streamable-http, auth none, reachable true, uptime30d 1) ## Packages _none_ ## Tools (34) - `agent_interact_gate`: Decide whether to INTERACT with an AI agent at all — any interaction. BROADER THAN should_i_pay: covers conversations, task delegation, adding an agent to your loop and copying its data (not just payments). Returns INTERACT / CAUTION / AVOID + confidence + a recommended recheck interval — designed to be re-checked before important interactions, not one-shot. Multichain ERC-8004 evidence + Solana. REST: GET /api/v1/agent-interact-gate ($0.02). Args: agent_id: ERC-8004 agent id (or use solana_asset) solana_asset: Solana 8004 agent pubkey (alternative) - `agent_liveness`: Is agent ALIVE right now? Verdict ALIVE/DEGRADED/DOWN/UNKNOWN before your agent wastes timeouts calling a dead service. CALL THIS BEFORE CALLING AN UNKNOWN AGENT'S ENDPOINT — a verdict backed by ~270 real probes/agent/30d (15-min probe cycle), NOT a one-shot ping like generic URL checkers. Returns action CALL/SLOW/AVOID + uptime % + p50/p95 latency + flaps + last-probe age. By ERC-8004 agent_id or Solana asset. REST: GET /api/v1/agent-liveness ($0.01). Args: agent_id: ERC-8004 agent id (or use solana_asset) solana_asset: Solana 8004 agent pubkey (alternative) window_days: SLA window 1-30 (default 7) - `check_agent_trust`: Decide whether to PAY or transact with an unknown agent: TRUST / CAUTION / UNTRUSTED / UNKNOWN verdict + reasons + 0-100 score. CALL THIS BEFORE PAYING OR INTEGRATING AN UNKNOWN AGENT — the pre-transaction firewall for the agent economy. Aggregates feedback across chains (Base + Ethereum + X Layer); per-chain breakdown is in signals.multichain. For Solana pass solana_asset (pubkey) instead of agent_id. An unknown agent NEVER returns TRUST (fail-closed). REST: GET /api/v1/agent/{agent_id}/trust-gate ($0.01). Args: agent_id: ERC-8004 agent id (or use solana_asset) chain_id: chain of the agent id (8453 Base, 1 Ethereum, 196 X Layer) solana_asset: Solana 8004 agent pubkey (alternative to agent_id) crosschain: aggregate feedback across chains (default true) - `check_base_token_safety`: Is this Base token BASED / MID / COOKED / RADIOACTIVE? Deterministic verdict before you buy: grade + action (ENTER_OK/CAUTION/SKIP) + confidence + red flags (honeypot, sell tax, mintable supply, owner privileges, holder concentration, thin liquidity) + expected loss for your amount. CALL THIS BEFORE SWAPPING INTO ANY BASE TOKEN. Fail-closed: unknown token returns UNKNOWN, never "safe". Risk screen, not investment advice. REST: GET|POST /api/v1/token/verdict ($0.02). Args: token: ERC-20 contract address on Base (0x…) amount_usd: planned position size in USD (scales expected_loss_usd) - `check_wallet_reputation`: Reputace a rizikovost adresy: LOW/MEDIUM/HIGH/UNKNOWN. Zdroj dat = on-chain registrace + indexované transakce. Ne SLA, ne úspěšnost obchodu. - `get_a2a_relationships`: A2A platební graf agentů: top hrany (kdo→komu, USDC), recurentní platby ('předplatné mezi stroji'), huby (cíle s mnoha plátci = infra uzly). - `get_agent_liveness`: Liveness agentů: kdo je alive/dormant/stale/dead/unknown — dle skutečné on-chain aktivity. - `get_agent_sla`: Liveness SLA jednoho agenta: uptime %, latence p50/p95, flapy — syntetický probing endpointů (historie roste od spuštění proberu). - `get_agent_trust_passport`: Full pre-payment dossier for one agent: ERC-8004 identity, sybil forensics, feedback-velocity class (organic/burst/farm) and 30d liveness compliance, with an ok / caution / avoid verdict. Use before integrating an agent or spending on it repeatedly — it bundles the trust-gate, forensics and SLA answers in one call. REST: GET /api/v1/agent/{agent_id}/trust-passport ($0.01). - `get_bazaar_growth_stats`: Denní vývoj x402 Bazaar katalogu: počet listingů, organické domény, payers/calls součty, nové/zaniklé listingy, odhad API objemu v USDC. - `get_bazaar_market_intel`: Market intelligence z denních Bazaar sweepů (30k+ listingů): kategorie dle reálné poptávky (payers) + nejrychleji rostoucí organické domény (den vs. den). Co agents REÁLNĚ kupují — data nikde jinde nejsou. - `get_bazaar_organic_ranking`: x402 Bazaar domény s ověřenou organickou poptávkou (calls/payers >= 3, >= 5 payers) z denního snapshotu katalogu — bez wallet farm. - `get_bittensor_network_health`: Živost Bittensor sítě (peers, head blok). Bez klíče — žádná subnety data. - `get_bittensor_subnet_trust`: Neuron trust řazení v subnetu (rank × incentive × stake, validator/miner, endpoint). - `get_bittensor_trust_overview`: Bittensor subnety dle emisí + trust proxy (activity, net flows). Top N default 20. - `get_btc_grid_signal`: Actionable BTC signal from our own live trading engine on Hyperliquid. Returns the engine's current directional signal, its active position with full detail (all levels, target, stop) and the live track record. Paper execution — engine decisions are real, PnL is paper with modeled fees, not investment advice. - `get_execution_health`: Execution health na Base: RPC latence, agent aktivita, gas trend. Base public RPC nemá mempool — žádná pending-tx data. - `get_pump_alerts`: Live pump.fun detection alerts with per-alert ACO decision. Args: minutes: lookback window (5-720) alert_type: optional filter — ignition | sudden_flow | pre_pump | whale_distribute | market_exit Returns recent alerts from our own pump.fun/Raydium behavioral collector (1350+ alerts/day): flow multiplier, smart buyer count and a decision per alert (WATCH_ENTER / EXIT_IF_LONG / AVOID). Not investment advice. - `get_reputation_summary`: ERC-8004 ReputationRegistry: počet feedbacků, agentů s reputací, unikátní klienti, průměrné skóre. - `get_signals`: Veřejně publikované signály (public delay vynucen). Data v raw token units; USD pouze u USDC. Podpis ověří nástroj verify_signal_proof. - `get_smart_money_feed`: LIVE smart-money feed for Solana meme trading — ONE payload, polling loop. Designed to sit INSIDE your trading decision loop (call every ~60s): fresh pump ignitions with SOL flow and unique-wallet counts, per-alert decision (WATCH_ENTER / EXIT_IF_LONG / STAND_ASIDE), smart buyer profiles (winrate / smart_score) and the top active whales right now. REST: GET /api/v1/alpha/smart-money ($0.02). Args: minutes: alert lookback window (5-720) limit: max alerts in the feed (1-50) Not investment advice. - `get_solana_agents`: Solana 8004 agents by ATOM trust quality (tier >= filter, default all). - `get_solana_forensics`: Solana sybil forensika: reviewer concentration, ghost ratio, burst, flags. - `get_solana_trust_summary`: 8004-Solana (mainnet) trust summary: agents, feedbacks, trust tier distribution. - `get_top_smart_buyers`: Smart buyers feed: wallety s nejlepším winrate × objemem (smart_score >= min_score). Kdo kupuje PŘED pumpou — zdroj copy-trading signálů. - `get_trading_decision`: CALL THIS BEFORE ENTERING CRYPTO PERPS: Actionable execution verdict. Trading Decision per asset from our own live Hyperliquid engines. Args: asset: btc | eth | hype | sol | uni | zec | pine_btc Returns an ACO verdict: ENTER_LONG / ENTER_SHORT / WAIT + entry reference + exit plan (target/stop) + live track record. Paper execution — engine decisions are real, PnL is paper with modeled fees, not investment advice. - `get_trending_agents`: Nejaktivnější on-chain autonomní agenti na Base, seřazení podle objemu (USDC transakce) a počtu transakcí. Limit max 20. - `get_trust_forensics`: Sybil forensika ERC-8004 feedbacků: rozdělení dle rizika, per-agent koncentrace reviewerů (Herfindahl), ghost-reviewer ratio, burst detekce, miner-vouch vs earned segmentation. risk: high|medium|low filtr. - `get_wallet_profile`: Wallet intelligence profil (Solanа, pump.fun behavioral data): winrate, typ (whale/bot/retail), smart_score, objem, průměrné držení, serial-creator flag. 87k+ profilů. Než zkopíruješ obchod, ověř kdo za ním stojí. - `mint_rug_check`: Is this Solana token a rug or safe to buy? Binary safety verdict — call it BEFORE buying or paying for any Solana meme token. Returns verdict SAFE/RISKY/RUG/UNKNOWN + action ENTER_OK/CAUTION/SKIP + confidence + specific red flags (active mint/freeze authority, dev dump, sell pressure, whale distribution, thin liquidity) + measured signals (unique wallets, buy share, bonding-curve liquidity, graduation status). Built from our own pump.fun/Meteora collector: 12M+ swap events, 63k launchpads, 1.3M whale wallets. Unknown mint returns UNKNOWN — never "safe". REST: GET /api/v1/solana/mint/{mint}/rug-check ($0.02). Args: mint: Solana token mint address - `reputation_origin`: Is this agent's reputation EARNED or BOUGHT? Origin verdict with reasons. THE FOLLOW-UP TO should_i_pay — explains WHY the verdict is CAUTION. Returns ORGANIC/BURST/FARM/MIXED/NONE + evidence: unique clients, burst-per-hour, ghost-wallet ratio, vouch-vs-earned feedback split. 90%+ of Base ERC-8004 reviewers show coordinated sybil behavior (arXiv 2606.26028) — this tool flags which agents' ratings are manufactured. By ERC-8004 agent_id or Solana asset. REST: GET /api/v1/reputation-origin ($0.02). Args: agent_id: ERC-8004 agent id (or use solana_asset) solana_asset: Solana 8004 agent pubkey (alternative) - `should_i_pay`: Pre-payment DECISION for any A2A payment: "should I pay agent X $N?". CALL THIS BEFORE EVERY PAYMENT TO AN UNKNOWN AGENT. The agent-facing equivalent of a credit check: pay $0.01 to avoid paying a scammer $N. Answers with action PAY / CAUTION / SKIP + confidence + expected loss in USDC sized to your exact amount, backed by on-chain ERC-8004 evidence (feedbacks, sybil forensics, liveness SLA). Multichain incl. Solana. Also available as REST: GET /api/v1/should-i-pay. Args: agent_id: ERC-8004 agent id of the payment TARGET (or use solana_asset) solana_asset: Solana 8004 agent pubkey (alternative to agent_id) amount_usd: the USDC amount you are about to pay Not investment advice. - `verify_signal_proof`: Ověří Proof-of-Signal (ed25519) proti payloadu. Vrací valid + key info. - `x402_seller_verdict`: Is this x402 seller a real business with paying users, or a farm/dead endpoint? Call it BEFORE your agent pays an unknown x402 endpoint. Returns verdict REAL/EMERGING/FARM/DEAD/UNKNOWN + action USE/CAUTION/AVOID + is_real (bool) + confidence + payers_30d, calls_30d, repeat-usage ratio (calls/payer) and the 11-day demand trend. Farms look like thousands of one-shot payers (calls/payer ~1.0); real businesses show repeat usage (calls/payer >= 3). Built from our daily sweep of the entire x402 catalog (32k+ listings/day, 11 days of history). Unknown seller returns UNKNOWN. REST: GET /api/v1/bazaar/seller-verdict ($0.02). Args: resource: full x402 resource URL (e.g. https://api.example.com/v1/foo) domain: or just a domain (e.g. api.example.com)