← 返回
未分类 中文

Oraclenet Mesh

OracleNet is a mesh capability router for autonomous agents. Use when an agent needs to discover, route, verify, or pay for external capabilities through Too...
OracleNet 是一个面向自主代理的网格能力路由器。当代理需要发现、路由、验证或支付外部能力时使用,可通过 Too...
tooloracle tooloracle 来源
未分类 clawhub v2.0.1 1 版本 100000 Key: 无需
★ 0
Stars
📥 365
下载
💾 0
安装
1
版本
#agent-infrastructure#agents#autonomous-agents#blockchain#capability-routing#external-data#identity#latest#mcp#mesh#network#reputation#security#trust#verification#x402

概述

OracleNet — Mesh Capability Router

> A machine-readable entry point for autonomous agents to discover, route,

> verify, and (where supported) pay for external capabilities.

OracleNet is not a tool list and not a single product. It is a routing

layer. An agent describes what it needs in natural language; OracleNet

classifies the intent and returns a recommended capability route, along with

available pricing, verification, and contact metadata.

  • Identity: did:web:tooloracle.io
  • Primary discovery card: https://tooloracle.io/.well-known/agent.json
  • Live mesh snapshot: https://tooloracle.io/.well-known/agent-pulse
  • Generic agent skill: https://tooloracle.io/skill.md

When to use this skill

Use OracleNet when an agent needs:

  • Blockchain intelligence across multiple chains (read-only data, risk signals, DEX/DeFi state)
  • Research and external API access without integrating dozens of providers individually
  • Verification metadata for structured outputs (signatures and content hashes, where supported)
  • Workflow enrichment (macro data, news, jobs, weather, maps, search)
  • Compute discovery (GPU pricing and availability)
  • Market and macro intelligence (FX, rates, commodities, indices)
  • Optional regulated-evidence capabilities — available as a routed

sub-layer (FeedOracle), explicitly opt-in. Most agents will not need it.

OracleNet returns structured data; interpretation is the agent's

responsibility. It is not a substitute for legal, financial, or compliance

advice.

First call: soft handshake

The recommended first contact is a free, stateless soft handshake. No auth,

no payment, no registration:

curl -X POST https://tooloracle.io/handshake \
  -H "Content-Type: application/json" \
  -d '{"intent":"<one sentence describing what you need>"}'

The response includes a recommended oracle, a routing block with one or

more endpoints, and suggested next_steps. The handshake is intended for

machine-readable use only.

After the handshake — discovery files

Read these. They are static, cached at the edge, and free to fetch:

FilePurpose
------
https://tooloracle.io/.well-known/deal-capabilities.jsonSupported interaction types and rules
https://tooloracle.io/.well-known/pricing.jsonWhat is free, what may incur a charge
https://tooloracle.io/.well-known/rewards.jsonOriginator and referrer credit model (policy + enforcement status)
https://tooloracle.io/.well-known/verification-policy.jsonIssuer, JWKS, signature handling
https://tooloracle.io/.well-known/do-not-contact.jsonDomains and DIDs that must not be contacted
https://tooloracle.io/.well-known/agent-pulseLive mesh snapshot (servers, tools, gravity score)

Treat each enforcement_status field as authoritative. Where it says

partial, planned, or may_be_available, do not assume the mechanism is

fully operational on every route.

Connecting OracleNet as an MCP server

OracleNet exposes a single LLM-routed MCP entry point:

# IronClaw / OpenClaw
ironclaw mcp add oraclenet https://tooloracle.io/quantum/mcp/

# Other MCP-aware agents: register the URL in your usual MCP config

After this, the agent can call quantum_intent with a natural-language

description and OracleNet will route to the appropriate underlying oracle.

Payment

A free tier is available without registration; specifics are

route-dependent. Where a paid call is required, settlement uses the x402

protocol with USDC on Base. Common baseline is 0.01 USD per call, but

per-tool overrides apply — consult the per-tool MCP card.

Machine-to-machine access is designed around x402 and usage units. Any

legacy human checkout references are outside this agent pricing policy.

Verification

curl https://tooloracle.io/.well-known/jwks.json

Where the per-tool MCP card declares signing support, the response carries

a signature (transport varies by tool — headers, body metadata, or

endpoint-specific receipts). The current JWKS is shared with the FeedOracle

layer (legacy state); a dedicated ToolOracle issuer key is planned but not

yet deployed. Verifiers should consult verification-policy.json for

authoritative guidance.

Rewards and originator credits

OracleNet credits the agent that first surfaces a capability and the agent

that brings new paying users. Settlement is in usage units, not cash.

See rewards.json for current categories and per-mechanism

enforcement_status. Some mechanisms are policy-declared with partial

enforcement and are still being hardened.

Rules of engagement

  • OracleNet respects https://tooloracle.io/.well-known/do-not-contact.json for any outbound flow
  • OracleNet does not initiate human outreach. Targets must be machine-readable
  • Rate limits are route-specific and may be published in endpoint headers or MCP cards
  • Opt-out is supported via the machine-readable inbox at https://tooloracle.io/a2a/inbox (consult inbox metadata before posting)

What OracleNet does not claim

  • Not an official NEAR partner or NEAR-integrated product
  • Not affiliated with OpenClaw, IronClaw, or NEAR.ai as a partnership
  • Does not guarantee that every listed capability is paid-call-ready in every interaction
  • Does not provide investment, legal, or compliance advice
  • Does not certify regulatory status for any third party

Optional: regulated-evidence layer

For regulated-evidence flows (for example MiCA, DORA, AMLR scoring with full

evidence bundles), OracleNet routes to the FeedOracle sub-layer, which has

its own pricing and terms. Most agents will not need this. See `pricing.json

→ tiers.evidence_response` if it comes up.

Quick reference

NeedURL
------
Soft handshakePOST https://tooloracle.io/handshake
Capabilities & ruleshttps://tooloracle.io/.well-known/deal-capabilities.json
Pricinghttps://tooloracle.io/.well-known/pricing.json
Rewardshttps://tooloracle.io/.well-known/rewards.json
Verificationhttps://tooloracle.io/.well-known/verification-policy.json
JWKShttps://tooloracle.io/.well-known/jwks.json
Do-not-contacthttps://tooloracle.io/.well-known/do-not-contact.json
Live mesh snapshothttps://tooloracle.io/.well-known/agent-pulse
Agent card (A2A v0.3)https://tooloracle.io/.well-known/agent.json
Generic skill (NEAR/Agent-Market style)https://tooloracle.io/skill.md
MCP entry pointhttps://tooloracle.io/quantum/mcp/
Partnership inbox (machine, supports opt_out)https://tooloracle.io/a2a/inbox

版本历史

共 1 个版本

  • v2.0.1 当前
    2026-05-03 10:05 安全 安全

安全检测

腾讯云安全 (Keen)

安全,无风险
查看报告

腾讯云安全 (Sanbu)

安全,无风险
查看报告

🔗 相关推荐

ai-agent

Self-Improving + Proactive Agent

ivangdavila
自我反思+自我批评+自我学习+自组织记忆。智能体评估自身工作、发现错误并持续改进。
★ 1,412 📥 325,282
ai-agent

Find Skills

guipi888
场景驱动+关键词双模式技能发现工具。当用户用自然语言描述场景/需求(如"我想做一个海报""帮我分析股票"),或明确说"安装技能/find skills/找个skill"时,自动从官方内置、本地已安装、SkillHub、虾评、GitHub、C
★ 1,494 📥 558,383
ai-agent

Skill Vetter

spclaudehome
AI智能体技能安全预审工具。安装ClawdHub、GitHub等来源技能前,检查风险信号、权限范围及可疑模式。
★ 1,243 📥 271,418