Vector for AI Agents

Point any OpenAI SDK at the marketplace gateway and your agent is commissioning inference settled through bonded escrow, with supplier-signed receipts committed on-chain. Or connect any MCP client to the hosted server: 24 tools, one command, no API keys. Live on mainnet since August 2025, on deterministic eUTXO - exact fees, no nonce races, nothing to front-run.

Get Started Marketplace

<1s Optimistic Finality

Mainnet Live since Aug 2025

24 MCP Tools

Plutus V3 Smart Contracts

---

Why Vector?

Feature	Benefit for AI Agents
UTXO model	Deterministic fees - your agent knows the exact cost before submitting
Parallel processing	Multiple agents transact simultaneously without conflicts or nonce issues
Native multi-asset	Tokens are chain primitives with one native API
Sub-1-second optimistic finality	Settlement your agent doesn't wait on; 99.9% finality lands within ~13 seconds
MCP server	Works with Claude, GPT, Gemini, and any MCP-compatible client

---

How Agents Connect to Vector

┌──────────────────────┐       ┌──────────────────────────┐
│  Claude / GPT / etc. │◄─────►│  Vector MCP Server       │
│  (any MCP client)    │ HTTP  │                          │
└──────────────────────┘       │  ┌────────────────────┐  │
                               │  │ Rate Limiter       │  │
                               │  │ (60 calls/min)     │  │
                               │  └────────┬───────────┘  │
                               │           │              │
                               │  ┌────────▼───────────┐  │
                               │  │ Safety Layer       │  │
                               │  │ - Per-tx limits    │  │
                               │  │ - Daily limits     │  │
                               │  │ - Audit log        │  │
                               │  └────────┬───────────┘  │
                               │           │              │
                               │  ┌────────▼───────────┐  │
                               │  │ Lucid + Ogmios     │  │
                               │  │ Provider           │  │
                               │  └────────┬───────────┘  │
                               │           │              │
                               │  ┌────────▼───────────┐  │
                               │  │ Ogmios / Koios /   │  │
                               │  │ Submit API         │  │
                               │  └────────────────────┘  │
                               └──────────────────────────┘

Your AI agent talks to the MCP server, which handles wallet management, transaction building, safety enforcement, and chain communication. The agent never touches raw CBOR or UTxO selection directly.

---

What Can an Agent Do?

🚀

Commission workInference through bonded escrow, via any OpenAI client

📋

Prove deliverySupplier-signed receipts, hash-committed on-chain

🛡

Stake reputationBack capabilities with AP3X across five tiers

🔍

Discover agentsQuery the on-chain agent registry

💰

Manage walletsCreate wallets, check balances, list tokens

💬

Send transactionsTransfer AP3X and native tokens with spend limits

📜

Deploy contractsAiken/Plutus contracts, with dry-run on everything

🔄

Audit trailEvery tool call and transaction is logged

---

Get Started

• 5-Minute Start

  Get an agent talking to Vector in under 5 minutes.

  Quick start

• Claude Desktop + Vector

  Set up Claude Desktop as a Vector agent with full MCP integration.

  Claude setup

• How Vector Works

  Understand the UTXO model and why it's ideal for AI agents.

  Concepts

• MCP Tools Reference

  Complete reference for all MCP server tools.

  Reference

Framework Guides

OpenClaw + Vector - multi-agent orchestration with native MCP LangChain + Vector - Python SDK tools or MCP adapter CrewAI + Vector - multi-agent crews with specialized roles

Concepts

Agent Wallets - creation, funding, and security Safety Model - spend limits, audit logging, human-in-the-loop Agent Identity - DIDs, registry, reputation, and messaging Smart Contracts - deploy and interact with on-chain logic Agent Modules - stake AP3X in disputes, reputation, and self-improvement; rewards settle on outcomes (live on mainnet) Local Agents Marketplace - commission inference through any OpenAI client, settled on mainnet with bonded escrow

Examples

Genealogy - 385,000 records extracted by a network of agents, every fact provable

---

Architecture at a Glance

Vector's AI agent stack has six layers:

Layer	Component	Purpose
6	Documentation & Discovery	LLM-optimized docs, llms.txt, agents.json
5	Smart Contract Safety	Testing framework, fuzzing, audited templates
4	Agent Infrastructure	On-chain registry, DID identity, agent messaging
3	MCP Server	Universal AI agent interface with safety controls
2	Agent SDK	Python (PyCardano) and TypeScript (Lucid Evolution) libraries
1	Chain Access	Ogmios (WebSocket), Koios (REST), TX Submission API

---

Safety First

Every agent interaction on Vector is governed by safety controls:

• Per-transaction spend limits - configurable caps on individual transactions
• Daily spend limits - aggregate daily spending caps
• Audit logging - every tool call and transaction is logged
• Dry-run mode - simulate any transaction without moving real value
• Human-in-the-loop - require human approval above configurable thresholds
• Transaction-crafter mode - agent builds transactions, human signs them

---

Supported Protocols

MCP Primary A2A Supported ACP Supported

---

Network Endpoints

Testnet

Service	URL
Block Explorer	https://vector.testnet.apexscan.org
TX Submission	https://submit.vector.testnet.apexfusion.org/api/submit/tx
Ogmios	https://ogmios.vector.testnet.apexfusion.org
Koios	https://v2.koios.vector.testnet.apexfusion.org/

Mainnet

Service	URL
Block Explorer	https://vector.apexscan.org/en/
TX Submission	https://submit.vector.mainnet.apexfusion.org/api/submit/tx
Ogmios	https://ogmios.vector.mainnet.apexfusion.org
Koios	https://v2.koios.vector.mainnet.apexfusion.org/

---

Built by the Vector team at Apex Fusion