Kronova Asset Intelligence Platform
A self-hosted orchestration platform for multi-agent workflows, real-world asset data, and secure settlement integrations.
Per-dimension scores and reasoning
The publishing workflows restrict permissions to contents:read and id-token:write and use repository secrets plus npm provenance, while the README sketches a separation between local orchestration and external settlement. However, the evidence does not document runtime tool permissions, user-confirmation gates, detailed data destinations, secret protection, log redaction, authorization for consequential actions, or rollback. Many dependencies use latest, installation uses npm install, and no lock, audit, or vulnerability-response evidence is supplied. Product naming, copyright, and licensing provide source attribution, but publisher identity remains unverified and the README clone path differs from the stated repository organization.
The README, package manifest, and publishing workflows consistently identify a Next.js/TypeScript project and provide basic build and release paths. Reliability credit is limited because the README calls most code production-worthy while also calling the platform and API a work in progress, and its architecture section contains duplicated, misspelled, and fragmentary text. Numerous floating dependencies and reliance on external or proprietary services weaken availability predictability. No evidence describes failure messages, graceful degradation, or recovery handling.
The README identifies developers, individuals, companies, and institutional users and outlines local development, agent orchestration, RWA, and production-settlement scenarios. It gives a high-level division between repository orchestration and AetherNet execution, but does not precisely enumerate supported and unsupported capabilities, agent trigger conditions, or enforcement boundaries. Environment guidance covers only a generic npm development path, without supported operating systems, an application Node version, a complete environment-variable reference, or a deployment compatibility matrix.
The README is divided into architecture, getting-started, production, contribution, and licensing sections and includes minimal installation commands. It lacks a complete configuration reference, working examples, an FAQ, and troubleshooting material. Naming is unstable across Kronova, AetherNet, QUAS, KVS, and differing GitHub organization paths. The README acknowledges ongoing development and distinguishes sandbox from production, but limitations are not catalogued systematically. The complete Apache-2.0 text, copyright notice, and trademark terms justify full license credit. A package version and tag-based SDK publishing exist, but no changelog or compatibility policy is shown. Contribution and enterprise-support paths are mentioned, although no verifiable maintainer or durable update channel is identified.
The material describes agent outputs, RAG, datasets, an RWA schema, billing, and many integrations, indicating potentially broad utility. Credit is limited because there are no output samples, end-to-end workflow examples, measured results, or comparative evidence demonstrating directly usable outputs or incremental value over alternatives. Local use is said not to require an AetherNet subscription, but the costs and benefits of Stripe billing, enterprise pilots, external models, and production settlement are unspecified.
The package manifest supports some stack and version claims, the workflows support SDK compilation, tag publishing, and npm-provenance claims, and the license file verifies licensing statements. Core claims about a 44-field schema, more than 30 models, 18 tools, production scalability, post-quantum security, deterministic settlement, and attack elimination are not traced to supplied code, tests, specifications, or independent materials. The README presents speculative and promotional security conclusions as facts, so fact/inference separation is absent.
- Do not treat README statements such as production-grade, mathematical attack elimination, or post-quantum security as verified guarantees; the supplied evidence contains no supporting implementation, tests, or specifications.
- Before connecting financial settlement, trading, refunds, escrow, or asset tokenization, require explicit human-confirmation controls, authorization scopes, audit logging, failure recovery, and rollback design.
- Pin and audit dependencies before deployment; widespread latest ranges and npm install create unpredictable supply-chain and compatibility changes.
- Verify ownership and update channels for the repository, npm SDK, maintainers, and AetherNet service because the publisher is not registry-verified and the repository paths are inconsistent.
- Obtain complete documentation for data flow, retention, encryption, deletion, and third-party sharing before handling OAuth tokens, API keys, payment information, customer data, or RWA records.
What does this agent do, and when should you use it?
This repository contains the open-source frontend and intelligence orchestration layer for the Kronova ecosystem, built with Next.js 16, TypeScript, Tailwind CSS, and Shadcn UI on a Supabase backend with a planned Qdrant migration. It lets users deploy, run, and continuously optimize multi-agent workflows across any AI model, with more than 30 models described as preconfigured. The platform combines embeddings, RAG, dataset management, 18 custom AI tools, OAuth 2.1 MCP server and client management, and Stripe subscriptions metered in ai_token units. Its asset workflow uses a 44-field RWA schema to create or import assets and feed agent output into reporting, learning, and optimization layers. The repository handles AI routing, data ingestion, and payload construction, while workflows that make constrained financial state changes are handed to the separate AetherNet QUAS API and Canton settlement layer. Local operation does not require an AetherNet subscription, but the README places secure production settlement, final RWA tokenization, and some institutional execution capabilities outside the open repository.
Users configure agents and complex workflows through the Next.js frontend, starting with preconfigured agents and flows or building their own. The orchestration layer selects and calls models, ingests data, manages embeddings, RAG, and datasets, and supports dataset imports from Qdrant and Weaviate; agents can invoke 18 custom AI tools, and their output can feed learning layers and perpetual ROI analytics. Asset workflows create or import RWA records using the 44-field asset schema and construct payloads for downstream execution. OAuth 2.1 MCP server and client management connects MCP participants, while the listed integrations include Salesforce, CrmOne, NetSuite, Shopify, Wix, Plaid, Resend, UPS, FedEx, Canton Network, and Sui. Production workflows requiring final settlement, conditional escrow, order or trade routing, or secure hardware interaction call the AetherNet QUAS API. The repository is therefore the open thinking and orchestration layer, not the complete secure settlement engine.
- An asset-technology team can host an RWA intake system, normalize records with the 44-field schema, and orchestrate agents that produce real-time asset intelligence reports.
- An enterprise automation team can connect multiple models, RAG datasets, and business systems in visual workflows, then route results into learning and optimization layers.
- An MCP platform developer can manage OAuth 2.1 MCP servers and clients while controlling tool or payload routing across multi-agent workflows.
- A SaaS product team can use Stripe ai_token metering to build usage-based subscriptions around model calls and agent workflows.
- An institutional pilot team can validate RWA tokenization or multiparty escrow workflows in a local sandbox before evaluating AetherNet QUAS and Canton for production settlement.
- A data team can import Qdrant or Weaviate datasets and use them in embedding, RAG, and agent workflows.
What are this agent's strengths and limitations?
- Combines multi-model orchestration, embeddings, RAG, dataset handling, MCP management, and asset workflows in one Next.js application.
- The 44-field RWA schema and real-time asset intelligence focus provide a more specific foundation for tokenization projects than a general-purpose workflow builder.
- Separates the open orchestration layer from high-risk settlement execution, allowing local development and testing without an AetherNet subscription.
- Includes Stripe ai_token metering and lists integrations spanning CRM, commerce, logistics, financial networks, and vector databases.
- Claims support for any AI model and includes more than 30 preconfigured models, reducing dependence on a single model provider.
- The README explicitly calls the platform and API a work in progress; adopters must independently validate stability, security, and scaling despite the claim that most code is near production-ready.
- Secure production settlement, Trusted Execution Environment access, and some institutional execution functions depend on the proprietary AetherNet QUAS API rather than this open repository alone.
- The current backend uses Supabase while Qdrant is described as a migration plan, creating possible future data-layer migration and operations work.
- The setup documentation omits an exact Node.js version, a complete environment-variable inventory, database bootstrap instructions, and production deployment steps.
- License evidence is inconsistent: repository metadata says NOASSERTION, while the README says Apache 2.0; adopters should verify the actual LICENSE file and trademark restrictions.
- The supplied material makes strong security and settlement claims but provides no tests, audits, benchmarks, or runnable QUAS API example that would substantiate them.
How do you install or deploy this agent?
The documented setup requires Git, npm, and a JavaScript runtime capable of running Next.js 16; no exact Node.js version is provided. Run:
git clone https://github.com/kronova/asset-intel-orchestration-engine.git
cd asset-intel-orchestration-engine
npm install
cp .env.example .env.local
npm run devFor local development, AetherNet variables may be left blank or set to the provided sandbox endpoints. The supplied material does not enumerate the remaining environment variables, Supabase bootstrap procedure, or a production deployment command, so adopters must inspect .env.example and the repository configuration.
How do you use this agent?
After installation, run npm run dev and use the local Next.js interface to configure models, agents, workflows, datasets, and asset records. Start from preconfigured agents and flows or build new ones, import data into an embedding or RAG pipeline, select tools and output learning layers, and run the workflow. Local orchestration and UI testing do not require an AetherNet subscription. Workflows involving final RWA tokenization, institutional MCP order or trade routing, multiparty conditional escrow, automated refund routing, off-chain data delivery, or Trusted Execution Environment interaction require enrollment in the enterprise pilot and AetherNet QUAS API credentials. No copyable QUAS API request is included in the supplied documentation.
How does this agent compare with similar options?
The README contrasts this Next.js 16, TypeScript, Tailwind CSS, Shadcn UI, and Supabase implementation with the separate kronova-rust-nextjs-example, aimed at developers who want a different interface and a Rust/rspc backend. It also distinguishes the local open orchestration sandbox from AetherNet QUAS: this repository performs model routing, ingestion, and payload construction, while QUAS handles trusted execution and Canton settlement.