Insight Logo
Insight
Oracle evidence
Price Query
APIAIVerifyDocumentation

Insight — Reference index

Evidence before execution.

Insight LogoInsight

Oracle transparency and risk infrastructure for DeFi. 15-minute price verification, cross-oracle comparison, position safety checks, and programmatic data access across 10+ providers and 40+ blockchain networks.

Platform

  • Home
  • Price Query
  • Price Insight
  • Oracle Directory
  • Daily Reports
  • API
  • Pricing
  • AI Agents

Safety

  • Safety Check
  • Pre-Trade Safety Check
  • Oracle Watch
  • Stablecoin Depeg
  • Wrapped Asset Peg

Resources

  • Documentation
  • API Reference
  • AI / MCP Docs
  • GitHub

© 2026 Insight. All rights reserved.

Privacy PolicyTerms of ServiceRefund PolicyContact
○

10 — Automate

AI × Crypto · An Oracle safety layer designed for machine execution

Try the Safety CheckConnect an Agent

Give every agent a reason before it acts.

Insight turns cross-oracle consensus, manipulation risk, freshness, and depeg exposure into machine-readable verdicts an agent can use to gate execution and explain its decision.

  • 01Pre-trade verdict
  • 02Continuous watch
  • 03Signed receipt
MCP shares the API credit meterSee pricingVerify a receiptMCP specification

01 — Decision surface

Built for the AI agent era

One platform that turns cross-oracle data into verifiable, agent-callable safety signals — the missing layer between AI agents and on-chain execution.

Pre-Trade Safety Check

Before any swap/borrow/lend, agents call one tool to verify cross-oracle consensus, deviation, freshness and depeg — getting a PASS / CAUTION / DANGER / BLOCK verdict.

Manipulation Risk Score

A 0–1 score blending per-provider deviation, cross-provider agreement, staleness and reputation — the single number an agent needs to decide if a price is trustworthy.

34 MCP Tools

Prices, consensus, risk summaries, position stress tests, stablecoin depeg, reputation rankings, feed health — all callable by any MCP-compatible client.

Agent-Native Verdicts

Every verdict ships with machine-readable factors AND a plain-English guidance line, so agents can both gate execution and explain decisions to humans.

Oracle Watch — the always-on companion

Pre-trade checks a single moment; Oracle Watch keeps watching. A live cross-oracle trust signal (NORMAL / CAUTION / DANGER) that agents can poll to keep running strategies — yield, keepers, portfolios — safe between trades.

Try it live →

02 — Gate the action

Flagship · Direction 1

Pre-Trade Oracle Safety Check

Try it live. Tell Insight what you're about to trade, and it aggregates cross-oracle consensus, deviation, freshness and depeg signals into one verdict an agent can act on.

$

Sign in or provide an API key below to run a live check.

Fill in the trade intent and run a check. Insight will aggregate cross-oracle consensus, deviation, freshness and depeg signals into a single safety verdict.

Same logic powers the pre_trade_safety_check MCP tool and the GET /api/v1/safety/pre-trade REST endpoint.

03 — Watch between actions

Always-on · Direction 2

Oracle Watch

Pre-trade checks the moment before an action. Oracle Watch keeps watching between actions — a live cross-oracle trust signal that running strategies can poll and gate on, instead of trusting a price silently.

Sign in or provide an API key below to run a live check.

Pick an asset and run Oracle Watch. Insight aggregates live cross-oracle deviation, agreement, quorum, outliers and staleness into one verdict an agent can gate on.

Call the MCP tool:
tool: oracle_watch
{
  "symbol": "ETH",
  "chain": "ethereum"
}

The agent returns a machine-readable verdict:
- Verdict: NORMAL | CAUTION | DANGER
- Recommendation: proceed | proceed_with_caution | halt
- Reason: within_tolerance | deviation_agreement_outlier_or_stale | deviation_or_agreement_breached_danger | insufficient_cross_oracle_quorum | ml_forward_risk_high | no_cross_oracle_coverage
Also gate on the composite Credibility Trust Score (0-100, low|medium|high).

For running strategies

Yield bots, keepers and portfolio managers poll the signal periodically and pause when it turns DANGER.

One verdict, same language

NORMAL / CAUTION / DANGER reuses the pre-trade severity thresholds (deviation 1%/3%, agreement 0.95/0.85), so agents speak one consistent risk language.

MCP + REST

Call it from any MCP client as oracle_watch, or hit GET /api/v1/oracle-watch directly. No cross-oracle coverage degrades to DANGER, not an error.

04 — Agent questions

What can your agent ask?

Once connected, ask in plain language — the agent picks the right MCP tool and returns the analysis, gating execution when risk is high.

Pre-trade gating

"I want to borrow 500k USDC against ETH on Aave — check the oracle is safe first, and abort if the verdict is DANGER or BLOCK."

Consensus & deviation

"What's the latest Chainlink price for BTC, and how far is it from the cross-oracle consensus?"

Automated risk briefing

"Generate an oracle health report for the last 24 hours, focusing on depeg events and feed latency."

Always-on monitoring

"I run a yield strategy on ETH. Poll oracle_watch every 5 minutes and pause withdrawals if the verdict ever turns DANGER."

05 — Connect the agent

Connect an AI agent in 3 steps

Any MCP-compatible client works. The agent auto-discovers Insight's tools — including the pre-trade safety check — and can start gating trades immediately.

1

Create an API Key

Sign in, generate an API key. API access is credit-metered — subscriptions from 49 USDC/mo or prepaid top-up packs.

2

Add the MCP server to your client

Copy the config for Claude Desktop, Cursor, or Windsurf and paste your key. One-click setup.

3

Gate every action and every running strategy

Agents auto-discover the pre_trade_safety_check tool for one-off trades and the oracle_watch tool to keep running strategies safe between actions.

Claude Desktopclaude_desktop_config.json
{
  "mcpServers": {
    "insight-oracle": {
      "url": "https://www.oracleinsight.xyz/api/mcp"
    }
  }
}
Cursor~/.cursor/mcp.json
{
  "mcpServers": {
    "insight-oracle": {
      "url": "https://www.oracleinsight.xyz/api/mcp"
    }
  }
}
Agent promptsystem prompt
Before executing any on-chain trade,
call pre_trade_safety_check with the
asset, chainId, action and amount.
If the verdict is DANGER or BLOCK,
do NOT execute — report the risk
factors to the user instead.
For long-running strategies, poll
oracle_watch periodically and pause
when the verdict turns DANGER.

06 — Generate the connection

Generate your MCP config

Pick your client, paste your API Key, and copy. Every request goes through the same authentication, rate limiting, and quota enforcement as the REST API.

Leave empty to generate a local stdio config; paste a key to generate a remote HTTP config.

MCP tool calls are credit-metered per call (C1–C4) — subscribe or top up credits. Every paying user gets all 34 tools. See pricing →

Cursor / Windsurf / Claude Desktop (HTTP)
{
  "mcpServers": {
    "insight-oracle": {
      "url": "https://oracleinsight.xyz/api/mcp"
    }
  }
}
Claude Desktop (local stdio)
{
  "mcpServers": {
    "insight-oracle": {
      "command": "npx",
      "args": [
        "tsx",
        "src/mcp/index.ts"
      ],
      "env": {
        "NODE_ENV": "production"
      }
    }
  }
}

07 — Keep the gate in code

REST integration

Generate integration code

Building a bot or script without MCP? Pick your language, fill in the trade intent, and copy a runnable snippet. Each one already embeds the immune-system gate — the trade aborts when the verdict is DANGER or BLOCK.

curl -X GET "https://oracleinsight.xyz/api/v1/safety/pre-trade?asset=ETH&chainId=1&action=swap&tradeAmountUsd=100000" \
  -H "X-API-Key: ins_xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx" \
  -H "Accept: application/json"

# Response: { "data": { "verdict": "PASS|CAUTION|DANGER|BLOCK", ... } }
# Immune-system gate — abort the trade on DANGER or BLOCK.

Every snippet embeds the immune-system gate: the trade is aborted when the verdict is DANGER or BLOCK. Same logic as the pre_trade_safety_check MCP tool.

08 — Control access

Sign in to manage API Keys

Sign in to create an API Key and have it auto-filled in the config generator and safety check demo.

09 — Test the tools

MCP Tool Playground

Test any of the 34 tools — including pre_trade_safety_check and oracle_watch — without leaving the browser.

Without signing in or providing an API Key, the Playground may fail due to missing authentication. Add a key in the config generator above, or sign in and retry.

Parameters

This tool takes no parameters.

Response

Call the tool on the left to see results

Commercial access

AI Agent Pricing

MCP and safety-check calls share the same credit meter and billing as the REST API. New users get 30 free trial credits after email verification — Developer from 49 USDC/mo (10,000 credits) or pay-as-you-go top-up packs.

View Pricing