What is Wielding.ai

Wielding.ai is an AI Agent platform. You create your own Agent, fill it with knowledge, configure how it talks, and publish it. Other people can then chat with your Agent — and pay per conversation in ATP (the platform's token).

There are two types of Agents: Human (your personal knowledge base) and Character (role-playing personas like historical figures or fictional characters). Both are backed by a structured knowledge system and powered by the LLM of your choice.

🎁

New accounts get 100 free ATP — enough to explore the platform and chat with published Agents right away.

Register & Create Your First Agent

Getting started takes about two minutes:

  1. Sign up — Click "Register" on the homepage. You'll get 100 ATP bonus automatically.
  2. Create an Agent — Go to your Dashboard and click "Create Agent". Pick a type: Human for a personal knowledge base, or Character for a role-playing persona.
  3. Add an LLM API key — In Agent Settings → LLM Config, add your API key for the model provider you want to use (DeepSeek, OpenAI, Anthropic, Doubao, or any OpenAI-compatible endpoint).
  4. Start chatting — Open the Chat tab and talk to your Agent. Knowledge gets extracted and stored automatically as you converse.

👤 Human vs 🎭 Character

Human agents are personal knowledge bases — they store facts, opinions, and research, and answer questions based on what you've taught them. Character agents have a configured persona (name, background, personality, speaking style) and respond in-character, grounded in their curated knowledge.

Add Knowledge

Your Agent is only as smart as its knowledge base. Here's how to fill it:

Method How
💬 Chat Just talk to your Agent. Facts are extracted and stored automatically from the conversation.
🔍 Research Use the /research [topic] command. The AI searches the web and saves relevant findings as structured facts.
✏️ Manual Add facts directly through the Knowledge panel — useful for precise, structured entries.
💡

For Character Agents, use /research to have the AI gather knowledge aligned with the character's identity. A "Feynman" agent can auto-research quantum physics, for example.

Chat Behavior Configuration

Every Agent has a Chat Config panel where you control how it responds. Here's what you can tweak:

Setting What it does
🗣️ Voice Profile Cognitive profile that shapes the Agent's personality and communication style.
🎭 Response Tone Choose from friendly, formal, academic, or role-play modes.
📐 Higher-Dimensional Perspective Agent occasionally adds meta-level insights from philosophy, systems thinking, or cognitive science.
🧠 Reasoning Chain Shows step-by-step reasoning when the answer involves inference across multiple knowledge entries.
⚡ Contradiction Detection Flags when retrieved knowledge contains conflicting information, instead of hiding the tension.
🌐 Web Enhancement When the knowledge base doesn't have enough info, the Agent searches the web to supplement its answer.

You can also fine-tune retrieval parameters: top_k (how many facts to retrieve), min_score (relevance threshold), max_tokens (response length), and context window (how many recent turns to include).

💬 Multi-Session Chat

Each Agent supports multiple conversation sessions. Create new sessions, switch between them, or delete old ones. The Agent remembers recent conversation turns within each session for context continuity. All responses stream in real-time.

Live2D Avatars

Give your Agent a visual presence with Live2D animated avatars that react during conversations.

  • 6 pre-built Live2D models to choose from
  • Upload your own custom model as a .zip file
  • Lip-sync animation plays automatically when the Agent is "speaking"

Configure avatars in Agent Settings → Live2D tab. The avatar appears in the chat interface when visitors talk to your Agent.

Publish & Pricing

By default, your Agent is private — only you can access it. To let others chat with it:

  1. Go to Agent Settings → Publish
  2. Set your per-chat ATP price (how much each conversation costs visitors)
  3. Toggle the Agent to "Public"

When someone chats with your published Agent, ATP is automatically deducted from their balance. You can track income and usage statistics in the Usage panel on your Dashboard.

💰

Set a price that reflects the value of your Agent's knowledge. A well-curated expert Agent can command higher ATP rates.

Three-Layer Knowledge Architecture

Every Agent's knowledge is organized in three layers that build on each other:

📝 Fact Layer — Storage

Raw knowledge fragments: notes, quotes, observations, data points. Stored in SQLite (metadata & full text) and Qdrant (semantic vectors for similarity search). This is the foundation — everything starts as a fact.

🔗 Relation Layer — Connections

Links between facts stored in a Neo4j knowledge graph. The AI automatically discovers relationships: cause-effect, similarity, contradiction, temporal sequence. This is where isolated facts become a web of knowledge.

🧠 Abstraction Layer — Insights

Higher-level patterns and insights generated from clusters of related facts. Think of it as the "so what?" layer — it surfaces themes, principles, and conclusions that emerge from your knowledge.

Visual Browsing

Explore your Agent's knowledge visually with three different views:

  • Graph View — Interactive force-directed graph. Facts are nodes, relations are edges. Zoom, pan, and click to explore how ideas connect.
  • Tree View — Hierarchical breakdown from abstractions down to supporting facts. See your knowledge structure at a glance.
  • Timeline — Chronological view of all entries. Track how your Agent's knowledge evolved over time.

Editing & Privacy

You have full control over your Agent's knowledge:

  • Edit any fact — update content, correct errors, add context
  • Delete facts you no longer want
  • Privacy controls — mark individual facts as private so they're used for reasoning but never shown to visitors

LLM Configuration

Each Agent needs an LLM to power its conversations. You bring your own API key. Supported providers:

Provider Popular Models Notes
Doubao Doubao Pro, Doubao Lite Great for Chinese-language Agents
DeepSeek DeepSeek Chat, DeepSeek Reasoner Strong reasoning, cost-effective
OpenAI GPT-4o, GPT-4o-mini Widely supported, versatile
Anthropic Claude Sonnet, Claude Opus Excellent for nuanced conversation
Custom Any model Any OpenAI-compatible API endpoint

Configure in Agent Settings → LLM Config. Each Agent can use a different provider and API key.

API Reference

Core API endpoints for programmatic access:

Method Endpoint Description
POST/api/auth/registerRegister a new account (auto-creates Agent + 100 ATP bonus)
POST/api/auth/loginLogin and get JWT token
GET/api/agentsList all your agents
POST/api/agentsCreate a new agent
GET/api/agents/{id}Get agent details
PUT/api/agents/{id}Update agent config
POST/api/agents/{id}/chatSend a chat message (streaming response)
GET/api/agents/{id}/knowledgeBrowse agent's knowledge base
GET/api/statsPlatform statistics
🔑

All endpoints (except register/login) require a Bearer token in the Authorization header. Get your token from the login response.