# `io.github.imokokok/insight-oracle` on Anlyon Radar

> Text inside fenced blocks marked untrusted-server-text was written by the MCP server's publisher, not by Anlyon Radar. It is data to inspect, not instructions to follow. Invisible characters are shown as [U+XXXX].
> No finding is not a safety guarantee: these are point-in-time checks of tool definitions, not of code or behaviour.

- Page: https://radar.anlyon.com/servers/io.github.imokokok/insight-oracle
- JSON: https://radar.anlyon.com/api/v1/servers/io.github.imokokok/insight-oracle
- Version: 0.3.1 (active)
- Publisher: io.github.imokokok (GitHub account)
- Runs: remote (streamable-http), local (npm)
- Sign-in to list tools: none
- Repository: https://github.com/imokokok/Insight
- Metadata from the official MCP Registry, synced 2026-11-06T14:18:22.876Z

Title and description, as the publisher wrote them:

```untrusted-server-text
Insight Pre-Trade Safety
Decision-input integrity for onchain AI agents: verify oracle prices before signing a trade.
```

Scanned 2026-11-06T13:18:24.102Z over streamable-http.

**81 tools: 9 do something that can't be undone, 25 change things and 47 only read.**

| Group | Class | Tools | Claude Code |
| --- | --- | --- | --- |
| Deletes or cancels | destructive | 0 | deny |
| Can't be undone | irreversible | 9 | ask |
| Changes things | write | 25 | ask |
| Only reads | read | 47 | allow |

## Exfil surface

No private-data -> untrusted-input -> outbound path in this server alone: no tool is tagged private-data or untrusted-input.

| Tag | Meaning | Tools |
| --- | --- | --- |
| private-data | Reads files, mail, messages, databases, secrets, calendars or repositories. | none |
| untrusted-input | Brings in text someone else can write: web pages, inbound email, messages, issues and comments. | none |
| outbound | Can carry data to people or systems: it sends, posts or publishes, or takes a recipient, channel or URL. | `mcp__insight-oracle__base_lending_guard_exact_call` `mcp__insight-oracle__base_perp_guard_exact_call` `mcp__insight-oracle__monad_lending_guard_exact_call` `mcp__insight-oracle__multichain_evm_lending_guard_exact_call` `mcp__insight-oracle__prepare_aerodrome_slipstream_swap` `mcp__insight-oracle__prepare_base_agent_swap_guard` `mcp__insight-oracle__prepare_pancakeswap_v3_swap` |

## Findings

No check matched anything in these tool definitions.


## Can't be undone (9)

Sends, pays, publishes or merges: once done, it cannot be taken back. Claude Code: ask.

### `base_lending_guard_exact_call`

- The name "base_lending_guard_exact_call" says "call": a word for a change.
- The input addresses people (input.transaction.to): what it sends cannot be taken back.
- Tag outbound: The input takes a destination (input.transaction.to): what the tool writes goes there.

Description, as the server wrote it:

```untrusted-server-text
Create portable Insight risk evidence for one exact unsigned direct Aave V3, Morpho Blue, Compound V3, or reviewed Moonwell Base supply/borrow call. It verifies target, selector, token or market binding, sender, nonce and zero value, then returns a short-lived EIP-712 decision when Insight signing is configured. It never authorizes, signs, or broadcasts the transaction; bind an ALLOW through PriorSeal or another user-authorization system.
```

### `base_perp_guard_exact_call`

- The name "base_perp_guard_exact_call" says "call": a word for a change.
- The input addresses people (input.transaction.to): what it sends cannot be taken back.
- Tag outbound: The input takes a destination (input.transaction.to): what the tool writes goes there.

Description, as the server wrote it:

```untrusted-server-text
Create portable Insight risk evidence for one exact unsigned direct Synthetix Perps V3 Base commitOrder call on a reviewed market (BTC, ETH, SOL, XRP, DOGE, PEPE; USDC margin). It verifies the canonical PerpsMarketProxy target, selector, reviewed market id, Pyth settlement strategy, sizeDelta direction, sender, nonce and zero value, then returns a short-lived EIP-712 decision when Insight signing is configured. It never authorizes, signs, or broadcasts the order; bind an ALLOW through PriorSeal or another user-authorization system.
```

### `monad_lending_guard_exact_call`

- The name "monad_lending_guard_exact_call" says "call": a word for a change.
- The input addresses people (input.transaction.to): what it sends cannot be taken back.
- Tag outbound: The input takes a destination (input.transaction.to): what the tool writes goes there.

Description, as the server wrote it:

```untrusted-server-text
Create portable Insight risk evidence for one exact unsigned direct Aave V3 Monad or Morpho Blue Monad supply/borrow call. It verifies target, selector, token or market binding, sender, nonce and zero value, then returns a short-lived EIP-712 decision when Insight signing is configured. It never authorizes, signs, or broadcasts the transaction; bind an ALLOW through PriorSeal or another user-authorization system.
```

### `multichain_evm_lending_guard_exact_call`

- The name "multichain_evm_lending_guard_exact_call" says "call": a word for a change.
- The input addresses people (input.transaction.to): what it sends cannot be taken back.
- Tag outbound: The input takes a destination (input.transaction.to): what the tool writes goes there.

Description, as the server wrote it:

```untrusted-server-text
Create portable Insight evidence for one exact unsigned direct reviewed EVM lending call on Base, Polygon, Arbitrum, or BNB Chain. It binds target, calldata, sender, nonce, value, protocol market, and short-lived Insight signature when configured. It never authorizes, signs, relays, or broadcasts the transaction; use PriorSeal for user authorization.
```

### `oracle_watch_window_begin`

- The description's first verb is "purchase": a word that usually means money moves, which cannot be taken back.

Description, as the server wrote it:

```untrusted-server-text
Purchase a 24h Oracle Watch monitored window for one asset (optionally scoped to one chain): ONE product price covers continuous halt/proceed manipulation-signal coverage, then poll the current signal inside the window at no per-call charge via oracle_watch_window_poll (budget: 300 polls, minimum 30s apart). This tool charges the full window product price once — it does not itself issue a signal. The response includes a one-time window token shown exactly once; keep it, every in-window call authenticates with it.
```

### `position_lifecycle_begin`

- The description's first verb is "purchase": a word that usually means money moves, which cannot be taken back.

Description, as the server wrote it:

```untrusted-server-text
Purchase a 24h Position Lifecycle Bundle for one position subject (optionally scoped to one chain): ONE product price covers an embedded Oracle Watch monitored window, 4 bundled guard assessments across base_perp, hyperliquid_perp, base_lending, monad_lending, multichain_evm_lending, or solana_lending, and 1 attested execution receipt. This tool charges the full bundle price once — it does not itself assess anything. The response includes a one-time lifecycle token shown exactly once; keep it, every bundled call authenticates with it (the embedded watch window accepts the SAME token).
```

### `prepare_aerodrome_slipstream_swap`

- The input addresses people (input.recipient): what it sends cannot be taken back.
- Tag outbound: The input takes a destination (input.recipient): what the tool writes goes there.

Description, as the server wrote it:

```untrusted-server-text
Prepare a user-signable exact-input swap against one allowlisted Base Aerodrome Slipstream pool. The tool takes a reviewed profile id, exact input amount, recipient, short deadline, and slippage bound; then obtains a pinned quote and re-verifies the Factory, QuoterV2, Pool, and Router identities. It returns unsigned approval and swap calldata only. Insight never receives a key, signs, relays, or broadcasts a transaction.
```

### `prepare_base_agent_swap_guard`

- The input addresses people (input.recipient): what it sends cannot be taken back.
- Tag outbound: The input takes a destination (input.recipient): what the tool writes goes there.

Description, as the server wrote it:

```untrusted-server-text
Prepare and gate one reviewed direct-router swap for an AI agent on Base, BNB Smart Chain, or Monad: Aerodrome Slipstream WETH→USDC CL100 and USDC→USDT CL1 (both directions) on Base, PancakeSwap V3 USDC→USDT 0.01% on Base, PancakeSwap V3 USDT→USDC 0.01% on BNB (both directions), or PancakeSwap V3 WMON→USDC 0.05% on Monad (both directions). It obtains a fresh pinned quote, re-verifies reviewed contract identities, canonical-decodes the exact approval/swap calldata, validates both agent decision prices against independent oracle consensus, and converts the route quote using the independent destination consensus before applying price-impact/TWAP/slippage/freshness policy. ALLOW returns exact unsigned approval and swap calldata plus a decisionHash, and adds a short-lived signed decision when Insight has a published attester; CAUTION or BLOCK returns no calls to sign. After settlement, pair the signed decision with the swap transaction hash via settle_base_agent_swap_guard to obtain a signed settlement closure. It does not route, sign, relay, broadcast, authorize, assess smart-account UserOperations, or accept arbitrary aggregator calldata.
```

### `prepare_pancakeswap_v3_swap`

- The input addresses people (input.recipient): what it sends cannot be taken back.
- Tag outbound: The input takes a destination (input.recipient): what the tool writes goes there.

Description, as the server wrote it:

```untrusted-server-text
Prepare a user-signable exact-input swap against one allowlisted Base PancakeSwap V3 pool. Insight obtains a pinned quote and re-verifies the Factory, QuoterV2, Pool, and Router runtime identities before returning unsigned approval and swap calldata. It never receives a key, signs, relays, or broadcasts a transaction.
```

## Changes things (25)

Creates or changes something that can be changed back. Claude Code: ask.

### `agent_begin_trade`

- Nothing in its name or description says what it does, so it is counted as a write.

Description, as the server wrote it:

```untrusted-server-text
Obtain oracle assessment evidence for a swap/DeFi intent BEFORE trading. Insight runs the oracle safety check and, if the verdict is PASS/CAUTION, returns a machine-readable "execution certification handle": preTradeUid, requestHash, CAIP-19 asset ids, the certified price (destination per source) and maxSlippageBps, plus a required operation-input check for both asset legs and a fresh Uniswap V3 on-chain route quote cross-checked by two Ethereum RPCs. If either input assessment is not SUFFICIENT, either route quote is stale or outside policy, or either oracle gate blocks, no execution handle is issued. The assessments are independently EIP-712 signed and linked to the receipt through both pre-trade UIDs. participant/source-group counts and the pre-trade signing time. Execute the trade with YOUR wallet, then call execution_receipt with this handle + txHash to obtain the signed, verifiable Execution Receipt. If the verdict is DANGER or BLOCK this tool refuses (do not trade). This is oracle assessment evidence, not principal authorization — pair it with execution_receipt.
```

### `assess_rwa_evidence`

- The description's first verb is "evaluate": a word for a change.

Description, as the server wrote it:

```untrusted-server-text
Evaluate supplied RWA/tokenized-stock evidence against an explicit policy. Unsigned diagnostic only: caller-supplied facts are not authenticated, and ALLOW does not authorize a trade. Existing pre-trade and authorization checks remain required.
```

### `check_liquidation_risk`

- Nothing in its name or description says what it does, so it is counted as a write.

Description, as the server wrote it:

```untrusted-server-text
Check protocol liquidation risk for a specific date. Returns stress-test results based on representative benchmark positions.
```

### `check_position_safety`

- Nothing in its name or description says what it does, so it is counted as a write.

Description, as the server wrote it:

```untrusted-server-text
Check the safety of a DeFi lending position against oracle deviation stress tests. Supports multi-asset or single-asset positions.
```

### `execution_preflight`

- Nothing in its name or description says what it does, so it is counted as a write.

Description, as the server wrote it:

```untrusted-server-text
Action-facing, chain-neutral data-sufficiency gate for an intended swap, borrow, lend, liquidation, or repay. It composes Insight’s strict multi-oracle pre-trade assessment with an optional fresh venue-quote assessment into ALLOW / CAUTION / BLOCK, lists the exact missing evidence, and returns the underlying signed assessment when available. It evaluates data sufficiency only — never contract safety, custody, profitability, or transaction execution. Monad is a first-class profile using on-chain Pyth Core evidence; set targetProviders to ["pyth"] for a low-latency native snapshot, which remains CAUTION under the strict quorum policy until independent evidence is present. Other supported chains use their configured oracle adapters.
```

### `execution_receipt`

- Nothing in its name or description says what it does, so it is counted as a write.

Description, as the server wrote it:

```untrusted-server-text
Issue signed execution-price evidence comparing an observed fill with the oracle-assessed quote and signed slippage band. Provide the pre-trade fields (preTradeUid, requestHash, asset ids, chain, oracle gate counts) plus the settlement txHash. Insight collects the on-chain fill, signs a receipt paired to the pre-trade via preTradeUid + requestHash, and returns a FAITHFUL / DEVIATED / NOT_EXECUTED / UNDETERMINED verdict. This reports supported price/fill fidelity, not principal authorization or overall transaction safety. Pair it with pre_trade_safety_check and oracle_watch.
```

### `hyperliquid_perp_guard_exact_call`

- The name "hyperliquid_perp_guard_exact_call" says "call": a word for a change.

Description, as the server wrote it:

```untrusted-server-text
Create portable Insight risk evidence for one exact unsigned Hyperliquid IOC order envelope on a reviewed market (BTC, ETH; USDC margin). It validates the single-order IOC envelope against the fresh assessment (side matches the assessed action, notional within the reviewed tolerance of the declared size, unexpired window, builder fee within the venue cap) and returns a short-lived EIP-712 decision over the envelope intent hash when Insight signing is configured. It never authorizes, signs, or submits the order; bind an ALLOW through PriorSeal or another user-authorization system, then sign the exact envelope in the user wallet process.
```

### `loopscale_vault_deposit`

- Nothing in its name or description says what it does, so it is counted as a write.

Description, as the server wrote it:

```untrusted-server-text
Build a single Solana mainnet Loopscale lending-vault deposit with Loopscale’s official production API, verify the live vault and principal mint, resolve every address lookup table, simulate the exact unsigned message, check oracle evidence, and return short-lived Insight evidence for PriorSeal. It does not sign, authorize, or broadcast. Deposit bundles requiring preliminary transactions are rejected rather than partially attested.
```

### `oracle_watch`

- Nothing in its name or description says what it does, so it is counted as a write.

Description, as the server wrote it:

```untrusted-server-text
Ongoing cross-oracle risk signal for an asset: current consensus deviation, agreement, quorum, outliers and staleness condensed into a NORMAL / CAUTION / DANGER verdict with a proceed / proceed_with_caution / halt recommendation. Agents should gate long-running strategies (yield, keeper, portfolio) on this signal, not just one-off trades. Pair it with pre_trade_safety_check for the decision moment.
```

### `oracle_watch_history`

- Nothing in its name or description says what it does, so it is counted as a write.

Description, as the server wrote it:

```untrusted-server-text
Retrospective Oracle Watch trend for an asset: verdict stability, degraded-time ratio, worst deviation and mean credibility over the last N days, plus the recent series. Use it to judge whether a feed has been dependable, not just whether it is healthy this instant. History is guaranteed only for the committed universe (ETH/BTC/USDC/USDT x ethereum/arbitrum/base); other pairs return a live point signal from oracle_watch but no curve.
```

### `oracle_watch_window_close`

- The name "oracle_watch_window_close" says "close": a word for a change.

Description, as the server wrote it:

```untrusted-server-text
Close an Oracle Watch monitored window early (idempotent; also works on an expired window) and return the summary of every signal issued inside it: poll counts, verdict distribution, and halt timestamps. Covered by the window purchase — no per-call charge. Authenticates with the one-time window token from the purchase.
```

### `oracle_watch_window_poll`

- Nothing in its name or description says what it does, so it is counted as a write.

Description, as the server wrote it:

```untrusted-server-text
Issue one current Oracle Watch halt/proceed signal inside an open monitored window purchased with oracle_watch_window_begin. Covered by the window purchase — no per-call charge — but counted against the window poll budget (300 per 24h, minimum 30s apart). Authenticates with the one-time window token from the purchase; without it the call fails closed.
```

### `position_lifecycle_close`

- The name "position_lifecycle_close" says "close": a word for a change.

Description, as the server wrote it:

```untrusted-server-text
Close a Position Lifecycle early (idempotent; also works on an expired lifecycle) and return the combined summary: guard and receipt usage plus the embedded watch window's coverage summary (polls, verdict distribution, halts). Covered by the lifecycle purchase — no per-call charge. Authenticates with the one-time lifecycle token.
```

### `position_lifecycle_guard`

- The description's first verb is "run": a word for a change.

Description, as the server wrote it:

```untrusted-server-text
Run one bundled guard assessment inside an open Position Lifecycle (purchased with position_lifecycle_begin). Choose the guard family per call; params use that family's native input (the same schema as the family's standalone guard tool), and solana_lending additionally requires the exact unsigned Solana lending transaction. Covered by the lifecycle purchase — no per-call charge — but consumes one of the 4 bundled guard slots; a BLOCK verdict is a complete assessment and consumes the slot. Authenticates with the one-time lifecycle token.
```

### `position_lifecycle_receipt`

- Nothing in its name or description says what it does, so it is counted as a write.

Description, as the server wrote it:

```untrusted-server-text
Issue the single bundled attested execution receipt inside an open Position Lifecycle (purchased with position_lifecycle_begin). The receipt fields are the same execution_receipt takes: the pre-trade fields the agent gated on plus the settlement txHash. Covered by the lifecycle purchase — no per-call charge — but the single bundled credit is consumed on success; a failed issuance never consumes it. Authenticates with the one-time lifecycle token.
```

### `pre_trade_safety_check`

- Nothing in its name or description says what it does, so it is counted as a write.

Description, as the server wrote it:

```untrusted-server-text
Pre-trade decision-input integrity checkpoint for AI agents. Call this BEFORE executing any on-chain swap/borrow/lend/liquidation/repay to verify that the price the action relies on can be trusted. First assessment domain: oracle risk — aggregates cross-oracle consensus prices, per-provider deviation, data freshness, stablecoin peg status, and reputation into a single verdict. An optional decisionInput object adds a separate Chainlink/CEX/DEX input-sufficiency assessment against the caller-supplied venue quote and explicit limits, with an independently verifiable EIP-712 signature when signing is enabled; it never changes the verdict. Verdicts are PASS / CAUTION / DANGER / BLOCK. AI agents MUST NOT execute trades when the verdict is DANGER or BLOCK. Also returns a recommended maximum position size. A PASS is not principal authorization or a guarantee of price correctness or economic safety.
```

### `prepare_hyperliquid_perp_order`

- Nothing in its name or description says what it does, so it is counted as a write.

Description, as the server wrote it:

```untrusted-server-text
Prepare a reviewed BTC, ETH, or HYPE Hyperliquid perpetual IOC order for LOCAL user/agent-wallet signing. Validates live asset metadata, lot/tick rules, caller notional cap, mark-oracle divergence, and displayed-L2 fill/impact bounds. Returns an unsigned, short-expiry exchange envelope only when every gate passes. Insight never creates a nonce, receives a key, signs, or submits an order.
```

### `recommend_oracle_setup`

- Nothing in its name or description says what it does, so it is counted as a write.

Description, as the server wrote it:

```untrusted-server-text
Recommend oracle provider setup for an asset based on active feeds. Use this to decide which providers to include in price feeds or risk analysis.
```

### `settle_base_agent_swap_guard`

- The description's first verb is "close": a word for a change.

Description, as the server wrote it:

```untrusted-server-text
Close the loop on a settled Base Agent Swap Guard swap. Present the signed guard decision the agent executed under plus the swap transaction hash (and the approval transaction hash when held). Insight verifies the decision, requires the on-chain to/value/calldata to match the guarded approval and swap calls byte-for-byte, reconciles the observed ERC-20 input and output amounts against the guarded quote (amountInRaw, quotedAmountOutRaw, amountOutMinimumRaw, realized output-leg slippage vs the agent policy), and returns a signed BaseAgentSwapGuardSettlement bound to the decisionHash with explicit reason codes (SWAP_TX_REVERTED, SOURCE_AMOUNT_MISMATCH, REALIZED_SLIPPAGE_EXCEEDED, DECISION_EXPIRED_AT_CLOSURE, APPROVAL_TX_NOT_PRESENTED, and others). A decision that does not verify, or a transaction whose bytes differ from the guarded calls, is rejected. This is settlement evidence, not transaction authority, a price-correctness claim, or a guarantee of good timing.
```

### `settle_solana_swap_guard`

- The description's first verb is "close": a word for a change.

Description, as the server wrote it:

```untrusted-server-text
Close the loop on a settled Solana Swap Guard swap. Present the signed guard decision the agent executed under plus the on-chain transaction signature. Insight verifies the decision, re-collects execution facts from Solana JSON-RPC, reconciles the observed SPL input and output balance deltas against the reviewed exact amounts (amountInRaw and the otherAmountThreshold minimum output), and returns a signed settlement bound to the decisionHash with an explicit status (SETTLED, REVERTED, or UNRECONCILED) and reconciliation code. A decision that does not verify is rejected. This is settlement evidence, not transaction authority, a price-correctness claim, or a guarantee of good timing.
```

### `solana_lending_guard`

- The description's first verb is "create": a word for a change.

Description, as the server wrote it:

```untrusted-server-text
Create portable Insight risk evidence for exactly one unsigned direct Kamino Lend, Marginfi v2, Jupiter Lend Earn deposit, or Save Lend reserve-liquidity deposit instruction on Solana mainnet. Save additionally verifies the submitted asset mint against the live source SPL token account. It accepts no transaction authority, verifies reviewed program-owned accounts, simulates the full unsigned message, checks both assets’ oracle evidence, and returns ALLOW / CAUTION / BLOCK. An ALLOW still needs a separately user-authorized PriorSeal binding before execution.
```

### `solana_swap_guard`

- Nothing in its name or description says what it does, so it is counted as a write.

Description, as the server wrote it:

```untrusted-server-text
Gate exactly one unsigned Orca Whirlpool swap on Solana mainnet: the reviewed USDC/USDT 0.01% pool (both directions). The caller supplies the fully resolved compiled unsigned transaction containing exactly one legacy Whirlpool swap instruction (no swapV2, no aggregators, no bundles). Insight canonical-decodes the instruction, re-verifies the pool, vaults, tick array, oracle, and both user token accounts from live mainnet state, prices the pool from its own sqrtPrice against independent oracle consensus, requires a positive otherAmountThreshold consistent with the declared decision prices, and simulates the exact unsigned message. ALLOW returns a short-lived signed decision bound to the exact message hash; CAUTION or BLOCK returns no executable payload. After settlement, pair the signed decision with the transaction signature via settle_solana_swap_guard. It does not sign, submit, relay, or authorize anything.
```

### `verify_base_agent_swap_guard_settlement`

- Nothing in its name or description says what it does, so it is counted as a write.

Description, as the server wrote it:

```untrusted-server-text
Independently verify a signed BaseAgentSwapGuardSettlement: structural shape, the reason-code hash, the EIP-712 uid, the attester signature, and attester trust. Pass the settlement artifact wrapped as { decision: { settlement: <settlement> } }. Returns valid/invalid with a machine-readable reason, plus the signed decisionHash, swapTxHash, and attester so a third party can detect two valid closures binding one decisionHash to different swap transactions — a contradiction no honest settlement produces. This checks the signature only; it does not re-read the chain.
```

### `verify_execution_pair`

- Nothing in its name or description says what it does, so it is counted as a write.

Description, as the server wrote it:

```untrusted-server-text
Verify that a pre-trade oracle-safety attestation and an Execution Receipt describe the same assessed trade and that the certify → execute → prove loop closed. Returns pairedValid, the closed-loop status (CLOSED_FAITHFUL / CLOSED_DEVIATED / CLOSED_NOT_EXECUTED / CLOSED_UNDETERMINED, PRICE_-prefixed on v3 receipts whose signed verdict is priceExecutionStatus, or PAIR_INVALID), and the binding assertions (preTradeUid, requestHash, destination gate + preTradeUidsHash on v3, chain, asset). On a v3 receipt that commits to a destination gate, pass destinationPreTradeAttestation or the pair cannot close. This checks assessment/price-evidence binding; it does not establish principal authorization, price correctness, or economic safety. Pair it with agent_begin_trade and execution_receipt.
```

### `verify_solana_swap_guard_settlement`

- Nothing in its name or description says what it does, so it is counted as a write.

Description, as the server wrote it:

```untrusted-server-text
Independently verify a signed Solana Swap Guard settlement: structural shape, the Ed25519 attester signature, and attester trust against the published Insight public key. Returns valid/invalid with a machine-readable reason plus the signed decisionHash, transaction signature, status, and attester so a third party can detect two valid closures binding one decisionHash to different transactions. This checks the signature only; it does not re-read the chain.
```

## Only reads (47)

Looks things up and changes nothing. Claude Code: allow.

### `base_lending_guard`

- The description's first verb is "read": a word for reading.

Description, as the server wrote it:

```untrusted-server-text
Read-only Base lending pre-signing gate for Aave V3, an exact Morpho Blue market, Compound V3 USDC Comet, or reviewed Moonwell markets. Separately evaluates collateral and loan oracle integrity, checks caller-declared post-action LTV against live protocol parameters, and returns ALLOW / CAUTION / BLOCK. It never signs, broadcasts, or authorizes a transaction; bind a successful decision to an exact user-authorized transaction separately.
```

### `base_perp_guard`

- The description's first verb is "read": a word for reading.

Description, as the server wrote it:

```untrusted-server-text
Read-only Base pre-trade decision-input integrity gate for opening a Synthetix Perps V3 position (reviewed BTC, ETH, SOL, XRP, DOGE, and PEPE markets, USDC margin). Separately evaluates underlying and margin oracle integrity, checks caller-declared post-action leverage against reviewed market caps, and returns ALLOW / CAUTION / BLOCK. When a reviewed Base spot market exists for the underlying (ETH today), it also attaches spot/TWAP/in-range liquidity evidence as disclosed context without entering oracle consensus. Assessment-only: it does not evaluate the venue mark-price or liquidation engine and never signs, broadcasts, or authorizes an order; bind a successful result to an exact user-authorized order separately.
```

### `compare_oracle_deviation`

- The name "compare_oracle_deviation" says "compare": a word for reading.

Description, as the server wrote it:

```untrusted-server-text
Compare historical price deviations across oracle providers for an asset. Helps identify which providers consistently diverge from consensus.
```

### `get_aerodrome_slipstream_market`

- The name "get_aerodrome_slipstream_market" says "get": a word for reading.

Description, as the server wrote it:

```untrusted-server-text
Get fresh, block-pinned market evidence for one allowlisted Aerodrome Slipstream Base pool. The pool’s Factory, QuoterV2, and runtime code identities are re-verified at the quoted block. This is DEX venue context, not an oracle-consensus price.
```

### `get_anomalies`

- The name "get_anomalies" says "get": a word for reading.

Description, as the server wrote it:

```untrusted-server-text
Get aggregated oracle anomaly events over a lookback period (1-30 days), including top deviation events and risk impacts.
```

### `get_base_dex_market`

- The name "get_base_dex_market" says "get": a word for reading.

Description, as the server wrote it:

```untrusted-server-text
Get fresh, block-pinned Base Uniswap V3 WETH/USDC market evidence: spot, 30-minute TWAP, active liquidity, and executable quote ladder. This is one DEX venue, not an oracle-consensus price.
```

### `get_base_dex_venue_integrations`

- The name "get_base_dex_venue_integrations" says "get": a word for reading.

Description, as the server wrote it:

```untrusted-server-text
Get block-pinned code-identity evidence for reviewed Base DEX deployments: Aerodrome Slipstream, PancakeSwap V3, Curve StableSwap, Maverick V2, QuickSwap V2 and Algebra V4, Uniswap V4, and PancakeSwap Infinity. The response distinguishes currently enabled unsigned-calldata lanes from venues that remain read-only until their pool-specific model, hook/plugin, and tests are reviewed.
```

### `get_consensus_price`

- The name "get_consensus_price" says "get": a word for reading.

Description, as the server wrote it:

```untrusted-server-text
Get the aggregated consensus price for an asset across multiple oracle providers. Useful when you need a robust, manipulation-resistant reference price.
```

### `get_correlation`

- The name "get_correlation" says "get": a word for reading.

Description, as the server wrote it:

```untrusted-server-text
Compute pairwise Pearson correlation of oracle provider deviations for an asset over a date range. Helps identify independent vs correlated providers.
```

### `get_coverage`

- The name "get_coverage" says "get": a word for reading.

Description, as the server wrote it:

```untrusted-server-text
Identify oracle coverage gaps and single-provider-risk symbols (assets served by only one oracle — a concentration risk). Returns breakdowns by provider, chain, and symbol. Use this when assessing where coverage is thin. For ecosystem scale plus top reputable providers, use get_metrics instead.
```

### `get_cross_chain_spreads`

- The name "get_cross_chain_spreads" says "get": a word for reading.

Description, as the server wrote it:

```untrusted-server-text
Get pairwise price spreads across chains for a given provider and symbol. Useful for cross-chain arbitrage and risk tracking.
```

### `get_daily_report`

- The name "get_daily_report" says "get": a word for reading.

Description, as the server wrote it:

```untrusted-server-text
Get the full daily oracle/risk report for a specific date, including liquidation risks, anomalies, and market summary.
```

### `get_feed_freshness`

- The name "get_feed_freshness" says "get": a word for reading.

Description, as the server wrote it:

```untrusted-server-text
Scan current freshness/staleness across many active feeds, filterable by provider, symbol, or category. Use this to find feeds needing attention right now. For a single feed by UUID use get_feed_health; for a dated ecosystem-wide report use get_oracle_health.
```

### `get_feed_health`

- The name "get_feed_health" says "get": a word for reading.

Description, as the server wrote it:

```untrusted-server-text
Get detailed point-in-time health status for ONE specific oracle feed by its UUID (consecutive failures, last success/failure, status). For ecosystem-wide health by date use get_oracle_health; to scan freshness across many feeds use get_feed_freshness.
```

### `get_feed_uptime`

- The name "get_feed_uptime" says "get": a word for reading.

Description, as the server wrote it:

```untrusted-server-text
Analyze feed data-delivery reliability over a date range: success rate, hours with data, coverage %, and average snapshots per day per provider+symbol. Use this to assess how completely a feed delivered data over time — distinct from get_feed_health (point-in-time status) and get_latency (speed).
```

### `get_feeds`

- The name "get_feeds" says "get": a word for reading.

Description, as the server wrote it:

```untrusted-server-text
List oracle feeds from the registry with optional filters (provider, symbol, category, chain, active status). Useful for discovering available feeds and their metadata.
```

### `get_hanji_monad_market`

- The name "get_hanji_monad_market" says "get": a word for reading.

Description, as the server wrote it:

```untrusted-server-text
Get fresh Hanji Monad MON/USDC CLOB context: code-pinned orderbook and fast-quoter proxy identities, proxy implementation identities, L2 spread, displayed depth, and bounded two-sided book-impact estimates. It is read-only venue context, not an oracle-consensus price or executable order quote.
```

### `get_hyperliquid_perp_market`

- The name "get_hyperliquid_perp_market" says "get": a word for reading.

Description, as the server wrote it:

```untrusted-server-text
Get fresh Hyperliquid mainnet perpetual-market evidence for a reviewed BTC, ETH, or HYPE market: mark/oracle divergence, funding, open interest, 24-hour notional volume, L2 spread, displayed depth, and two-sided book-impact estimates. It is read-only CLOB context, not an oracle-consensus price or executable exchange quote.
```

### `get_incidents`

- The name "get_incidents" says "get": a word for reading.

Description, as the server wrote it:

```untrusted-server-text
Get oracle incidents and deviation events within a date range. Filter by provider and minimum severity.
```

### `get_latency`

- The name "get_latency" says "get": a word for reading.

Description, as the server wrote it:

```untrusted-server-text
Analyze 15-minute oracle collector latency statistics (min, mean, p50, p90, p95, p99) over a date range, optionally filtered by provider or symbol.
```

### `get_metrics`

- The name "get_metrics" says "get": a word for reading.

Description, as the server wrote it:

```untrusted-server-text
Compact ecosystem overview: provider/feed/symbol/chain/category counts plus top providers ranked by reputation score. Use this for a quick scale snapshot or provider ranking. To find coverage gaps and single-provider risk, use get_coverage instead.
```

### `get_monad_dex_market`

- The name "get_monad_dex_market" says "get": a word for reading.

Description, as the server wrote it:

```untrusted-server-text
Get fresh, block-pinned Monad Uniswap V3 WMON/USDC market evidence: spot, 15-minute TWAP, active liquidity, executable quote ladder, and finalized rolling volume. This is a single-venue DEX market source, not an oracle-consensus price.
```

### `get_oracle_health`

- The name "get_oracle_health" says "get": a word for reading.

Description, as the server wrote it:

```untrusted-server-text
Get the ecosystem-wide oracle health report for a specific date (uptime, staleness, and success rates across all providers). Use this for a dated snapshot of overall health. For a single feed by UUID use get_feed_health; to scan current freshness across many feeds use get_feed_freshness.
```

### `get_oracle_price`

- The name "get_oracle_price" says "get": a word for reading.

Description, as the server wrote it:

```untrusted-server-text
Fetch the latest price from a specific oracle provider for an asset.
```

### `get_oracle_prices_batch`

- The name "get_oracle_prices_batch" says "get": a word for reading.

Description, as the server wrote it:

```untrusted-server-text
Fetch prices for multiple assets/providers in a single request. Up to 20 queries.
```

### `get_pancakeswap_v3_market`

- The name "get_pancakeswap_v3_market" says "get": a word for reading.

Description, as the server wrote it:

```untrusted-server-text
Get fresh, block-pinned market evidence for one allowlisted PancakeSwap V3 pool on Base or BNB Smart Chain. The Factory, QuoterV2, and Pool runtime identities are re-verified at the quoted block. This is DEX venue context, not an oracle-consensus price.
```

### `get_price_history`

- The name "get_price_history" says "get": a word for reading.

Description, as the server wrote it:

```untrusted-server-text
Get historical price data for an asset from a specific oracle provider over a period (in hours).
```

### `get_protocol_oracle_exposure`

- The name "get_protocol_oracle_exposure" says "get": a word for reading.

Description, as the server wrote it:

```untrusted-server-text
Analyze oracle provider concentration risk for a specific lending protocol. Shows which providers cover which assets and concentration metrics.
```

### `get_protocol_risk_params`

- The name "get_protocol_risk_params" says "get": a word for reading.

Description, as the server wrote it:

```untrusted-server-text
Get risk parameters for a specific DeFi protocol, including liquidation thresholds, LTV ratios, and collateral factors.
```

### `get_protocols`

- The name "get_protocols" says "get": a word for reading.

Description, as the server wrote it:

```untrusted-server-text
Get a list of lending protocols with dynamic data (TVL, asset counts, oracle providers used). Optionally filter by name/slug query.
```

### `get_provider_reputation`

- The name "get_provider_reputation" says "get": a word for reading.

Description, as the server wrote it:

```untrusted-server-text
Get detailed reputation metrics for a specific oracle provider, optionally with historical trend.
```

### `get_quickswap_v2_market`

- The name "get_quickswap_v2_market" says "get": a word for reading.

Description, as the server wrote it:

```untrusted-server-text
Get fresh, block-pinned Base QuickSwap V2 USDC/USDT market evidence: factory/router/pair identities, reserves, spot, and a bounded exact-input quote ladder. This is one reviewed pair, not an oracle-consensus price or a generic QuickSwap route.
```

### `get_reputation_rankings`

- The name "get_reputation_rankings" says "get": a word for reading.

Description, as the server wrote it:

```untrusted-server-text
Get oracle provider reputation rankings with trend over a period (1-90 days).
```

### `get_risk_summary`

- The name "get_risk_summary" says "get": a word for reading.

Description, as the server wrote it:

```untrusted-server-text
Get a composite risk summary for an asset across selected oracle providers. Includes HHI concentration, volatility, correlation, freshness, manipulation resistance, and shared-dependency risk.
```

### `get_robinhood_rwa_context`

- The name "get_robinhood_rwa_context" says "get": a word for reading.

Description, as the server wrote it:

```untrusted-server-text
Get Robinhood Stock Token issuer context: official asset identity, trading capabilities, halt state, corporate actions, multiplier-normalized reference price, and optional on-chain ERC-8056 verification. First-party evidence only; it is not an independent oracle and never counts toward oracle quorum or authorizes a trade.
```

### `get_robinhood_rwa_instrument`

- The name "get_robinhood_rwa_instrument" says "get": a word for reading.

Description, as the server wrote it:

```untrusted-server-text
Resolve a Robinhood Stock Token to Insight committed master data, using share-class FIGI for the underlying and ISO 10383 MIC for its primary market, then cross-check issuer UID, ISIN and contract deployment. Reference/admission evidence only; it never counts toward price quorum or authorizes a trade.
```

### `get_settlement_evidence`

- The name "get_settlement_evidence" says "get": a word for reading.

Description, as the server wrote it:

```untrusted-server-text
Get Insight-signed settlement evidence for prediction-market crypto price contracts: the declared outcome vs Insight’s independent verdict, price change and oracle deviation in basis points, candle coverage, and a consistency grade (consistent / inconsistent / insufficient_data). Pass slug for one market, or omit it to list recent packs. Derived analytics only; raw venue candles are not distributed.
```

### `get_solana_dex_route`

- The name "get_solana_dex_route" says "get": a word for reading.

Description, as the server wrote it:

```untrusted-server-text
Get fresh Jupiter SOL/USDC exact-input route evidence: quote ladder, price impact, selected AMM route, quote context slot, and finalized Solana RPC witness. It is a route quote, not a simulated or guaranteed swap.
```

### `get_stablecoin_list`

- The name "get_stablecoin_list" says "get": a word for reading.

Description, as the server wrote it:

```untrusted-server-text
List all tracked stablecoin symbols.
```

### `get_stablecoin_peg`

- The name "get_stablecoin_peg" says "get": a word for reading.

Description, as the server wrote it:

```untrusted-server-text
Check stablecoin peg status. Provide a symbol for a specific stablecoin, or omit it to get all tracked stablecoins.
```

### `get_symbols`

- The name "get_symbols" says "get": a word for reading.

Description, as the server wrote it:

```untrusted-server-text
List supported asset symbols. Optionally filter by a search query. Use this when you are unsure whether a symbol is supported.
```

### `get_wrapped_asset_peg`

- The name "get_wrapped_asset_peg" says "get": a word for reading.

Description, as the server wrote it:

```untrusted-server-text
Check wrapped asset (e.g. WBTC, wstETH) peg status. Provide a symbol for a specific asset, or omit it to get all tracked wrapped assets.
```

### `hyperliquid_perp_guard`

- The description's first verb is "read": a word for reading.

Description, as the server wrote it:

```untrusted-server-text
Read-only Hyperliquid pre-trade decision gate for opening a perpetual position on the Hyperliquid L1 venue (reviewed BTC and ETH markets today, USDC margin). Separately evaluates underlying and USDC cross-oracle consensus integrity on canonical mainnet feeds, checks caller-declared post-action leverage against a reviewed static profile capped by the freshly observed venue leverage limit, and attaches disclosed venue evidence (mark/oracle divergence, cross-venue consensus deviation, displayed depth) that can escalate the verdict. Returns ALLOW / CAUTION / BLOCK. Assessment-only: it does not evaluate the liquidation engine and never signs, submits, or authorizes an order; bind a successful result to an exact user-authorized order envelope separately.
```

### `monad_lending_guard`

- The description's first verb is "read": a word for reading.

Description, as the server wrote it:

```untrusted-server-text
Read-only Monad lending pre-signing gate for direct Aave V3 or an exact Morpho Blue market. Separately evaluates collateral and loan oracle integrity, checks caller-declared post-action LTV against live protocol parameters, and returns ALLOW / CAUTION / BLOCK. It never signs, broadcasts, or authorizes a transaction; bind a successful decision to an exact user-authorized transaction separately.
```

### `multichain_evm_lending_guard`

- The description's first verb is "read": a word for reading.

Description, as the server wrote it:

```untrusted-server-text
Read-only lending pre-signing gate for reviewed Base, Polygon, Arbitrum, and BNB Chain deployments: Aave V3, Venus Core, Compound V3, Dolomite, Morpho Blue, Lista Moolah, Euler v2, and Fluid T1. Euler borrowing is accepted only as the reviewed four-step EVC batch; Dolomite only as one sender-owned account with one direct deposit or borrow; Fluid T1 only as collateral-only supply into a factory-authenticated vault. The tool checks caller-declared post-action values against live protocol state and independent oracle evidence, then returns ALLOW / CAUTION / BLOCK. It never signs, broadcasts, or authorizes a transaction.
```

### `solana_perp_guard`

- The description's first verb is "read": a word for reading.

Description, as the server wrote it:

```untrusted-server-text
Read-only Solana pre-trade decision gate for opening a perpetual position on Drift Protocol v2 on Solana mainnet (reviewed SOL, BTC, and ETH perp markets today, USDC margin). Drift publishes no REST API, so venue evidence is read directly from Drift on-chain Anchor accounts through Solana mainnet RPC: mark price from the AMM curve, the market’s own Pyth Lazer oracle price, funding rate, open interest versus the on-chain maxOpenInterest backstop, and live margin ratios. Separately evaluates underlying and USDC cross-oracle consensus integrity on canonical mainnet feeds, checks caller-declared post-action leverage against a reviewed static profile capped by the live on-chain margin ratio, and attaches disclosed venue evidence that can escalate the verdict. Returns ALLOW / CAUTION / BLOCK. Assessment-only: it does not evaluate Drift’s liquidation engine and never signs, submits, or authorizes a transaction; bind a successful result to one exact user-authorized unsigned transaction separately.
```

### `verify_hyperliquid_perp_execution`

- The description's first verb is "read": a word for reading.

Description, as the server wrote it:

```untrusted-server-text
Read and verify the public Hyperliquid status and bounded fill history for one user-signed reviewed BTC, ETH, or HYPE perp order. Compares the returned order and fills to the expected client-order-id intent. This is an observed Info-API record, not an Insight signature, a wallet authorization, or a trade submission.
```

## Claude Code permissions

For `.claude/settings.json`, assuming the server is installed as `insight-oracle`:

```json
{
  "permissions": {
    "allow": [
      "mcp__insight-oracle__get_robinhood_rwa_instrument",
      "mcp__insight-oracle__get_robinhood_rwa_context",
      "mcp__insight-oracle__get_oracle_price",
      "mcp__insight-oracle__get_consensus_price",
      "mcp__insight-oracle__get_risk_summary",
      "mcp__insight-oracle__get_oracle_health",
      "mcp__insight-oracle__compare_oracle_deviation",
      "mcp__insight-oracle__get_stablecoin_peg",
      "mcp__insight-oracle__get_protocol_risk_params",
      "mcp__insight-oracle__get_symbols",
      "mcp__insight-oracle__get_stablecoin_list",
      "mcp__insight-oracle__get_oracle_prices_batch",
      "mcp__insight-oracle__get_price_history",
      "mcp__insight-oracle__get_cross_chain_spreads",
      "mcp__insight-oracle__get_wrapped_asset_peg",
      "mcp__insight-oracle__get_protocols",
      "mcp__insight-oracle__get_protocol_oracle_exposure",
      "mcp__insight-oracle__get_feed_freshness",
      "mcp__insight-oracle__get_feeds",
      "mcp__insight-oracle__get_feed_health",
      "mcp__insight-oracle__get_feed_uptime",
      "mcp__insight-oracle__get_reputation_rankings",
      "mcp__insight-oracle__get_provider_reputation",
      "mcp__insight-oracle__get_daily_report",
      "mcp__insight-oracle__get_incidents",
      "mcp__insight-oracle__get_settlement_evidence",
      "mcp__insight-oracle__get_latency",
      "mcp__insight-oracle__get_anomalies",
      "mcp__insight-oracle__get_correlation",
      "mcp__insight-oracle__get_coverage",
      "mcp__insight-oracle__get_metrics",
      "mcp__insight-oracle__get_monad_dex_market",
      "mcp__insight-oracle__get_hanji_monad_market",
      "mcp__insight-oracle__get_base_dex_market",
      "mcp__insight-oracle__get_base_dex_venue_integrations",
      "mcp__insight-oracle__get_aerodrome_slipstream_market",
      "mcp__insight-oracle__get_pancakeswap_v3_market",
      "mcp__insight-oracle__get_quickswap_v2_market",
      "mcp__insight-oracle__get_solana_dex_route",
      "mcp__insight-oracle__get_hyperliquid_perp_market",
      "mcp__insight-oracle__verify_hyperliquid_perp_execution",
      "mcp__insight-oracle__base_lending_guard",
      "mcp__insight-oracle__base_perp_guard",
      "mcp__insight-oracle__hyperliquid_perp_guard",
      "mcp__insight-oracle__monad_lending_guard",
      "mcp__insight-oracle__multichain_evm_lending_guard",
      "mcp__insight-oracle__solana_perp_guard"
    ],
    "ask": [
      "mcp__insight-oracle__assess_rwa_evidence",
      "mcp__insight-oracle__check_liquidation_risk",
      "mcp__insight-oracle__recommend_oracle_setup",
      "mcp__insight-oracle__prepare_aerodrome_slipstream_swap",
      "mcp__insight-oracle__prepare_base_agent_swap_guard",
      "mcp__insight-oracle__settle_base_agent_swap_guard",
      "mcp__insight-oracle__verify_base_agent_swap_guard_settlement",
      "mcp__insight-oracle__prepare_pancakeswap_v3_swap",
      "mcp__insight-oracle__prepare_hyperliquid_perp_order",
      "mcp__insight-oracle__base_lending_guard_exact_call",
      "mcp__insight-oracle__base_perp_guard_exact_call",
      "mcp__insight-oracle__hyperliquid_perp_guard_exact_call",
      "mcp__insight-oracle__monad_lending_guard_exact_call",
      "mcp__insight-oracle__multichain_evm_lending_guard_exact_call",
      "mcp__insight-oracle__solana_lending_guard",
      "mcp__insight-oracle__solana_swap_guard",
      "mcp__insight-oracle__settle_solana_swap_guard",
      "mcp__insight-oracle__verify_solana_swap_guard_settlement",
      "mcp__insight-oracle__loopscale_vault_deposit",
      "mcp__insight-oracle__check_position_safety",
      "mcp__insight-oracle__pre_trade_safety_check",
      "mcp__insight-oracle__oracle_watch",
      "mcp__insight-oracle__oracle_watch_history",
      "mcp__insight-oracle__oracle_watch_window_begin",
      "mcp__insight-oracle__oracle_watch_window_poll",
      "mcp__insight-oracle__oracle_watch_window_close",
      "mcp__insight-oracle__position_lifecycle_begin",
      "mcp__insight-oracle__position_lifecycle_guard",
      "mcp__insight-oracle__position_lifecycle_receipt",
      "mcp__insight-oracle__position_lifecycle_close",
      "mcp__insight-oracle__execution_receipt",
      "mcp__insight-oracle__execution_preflight",
      "mcp__insight-oracle__agent_begin_trade",
      "mcp__insight-oracle__verify_execution_pair"
    ],
    "deny": []
  }
}
```

Capability tags behind the rules (a read tagged outbound is asked about):

- `mcp__insight-oracle__prepare_aerodrome_slipstream_swap`: outbound
- `mcp__insight-oracle__prepare_base_agent_swap_guard`: outbound
- `mcp__insight-oracle__prepare_pancakeswap_v3_swap`: outbound
- `mcp__insight-oracle__base_lending_guard_exact_call`: outbound
- `mcp__insight-oracle__base_perp_guard_exact_call`: outbound
- `mcp__insight-oracle__monad_lending_guard_exact_call`: outbound
- `mcp__insight-oracle__multichain_evm_lending_guard_exact_call`: outbound

## Remote posture

What an anonymous client saw at the remote endpoint, last 2026-11-20T19:06:04.383Z, unchanged since 2026-11-20T19:06:04.383Z. No credentials were sent and no tool was called. Values are shown as the endpoint returned them. Method: https://radar.anlyon.com/methodology#remote-posture

| Fact | Observed |
| --- | --- |
| Reachable | `yes` |
| HTTP status of an anonymous initialize | `402` |
| tools/list served without credentials | `no` |
| Tools that change things, listed without credentials | `0` |
| Protocol version | `none` |
| WWW-Authenticate on 401 | `not observed` |
| resource_metadata in WWW-Authenticate | `none` |
| RFC 9728 protected resource metadata | `none` |
| Authorization servers | `none` |
| Legacy HTTP+SSE transport | `no` |
| Dynamic client registration | `not observed` |
| Client ID metadata documents | `not observed` |

Probe stopped: `initialize answered HTTP 402.`

Posture changes:

- 2026-11-20T19:06:04.383Z: HTTP status of an anonymous initialize `200` -> `402`; tools/list served without credentials `yes` -> `no`; Tools that change things, listed without credentials `15` -> `0`; Protocol version `2025-06-18` -> `none`
- 2026-11-13T12:04:36.230Z: Tools that change things, listed without credentials `34` -> `15`
- 2026-11-06T05:03:24.587Z: Tools that change things, listed without credentials `19` -> `34`
- 2026-10-29T22:02:29.597Z: Tools that change things, listed without credentials `10` -> `19`
