Character Builder
Character Builder
The Visual Builder is a canvas-based editor for creating and configuring AI characters. It combines visual node placement with detailed configuration panels.
Accessing the Builder
The Visual Builder is served by Qui Anima. You can access it through the Anima interface or from the QUI Core dashboard's Services section.
[Screenshot: Visual Builder overview showing canvas, palette, and character bar]
Interface Layout
The builder has three main areas:
Canvas (Center)
The main workspace. Your character's core node sits at the center, with capability nodes orbiting around it connected by gradient hierarchy lines. These lines show "this character has these capabilities" — they are visual indicators, not workflow connections.
- Left-click a node — opens its configuration panel
- Left-click and drag — moves the node; it snaps back to its orbital position on release
- Middle-click and drag — pans the entire canvas view
Node Palette (Left)
The palette contains all available node types organized into two tabs:
Basics tab:
- Attributes — Character definition (system prompts, personality) and Token Limits (input/output/context budgets)
- Abilities — M2M messaging, Self-Modify, Spark sub-agents, Agentic mode, and other core capabilities
Cognitive Design tab:
- Memory — Semantic memory configuration
- Working Memory — Clipboard (session-scoped) and Variable Store (persistent)
- Cortex — Memory consolidation processors and modes (flyout menus with 9 processors and 8 modes)
- Tools — Terminal, MCP tools (165+ across 9 categories), and integrations
- Thalamus — Event routing, triggers, perception nodes, and channel adapters
To add a node, click it in the palette. It appears on the canvas connected to the core node.
Character Bar (Bottom)
Shows all your saved characters as clickable tiles. Click a character to load it into the builder. Click [+] to create a new character.
Creating a Character
- Click [+] in the character bar at the bottom
- The new character appears with a core Anima node at the center
- Click the core node to configure the basics:
- Name — the character's display name
- Description — a brief summary of who this character is
- LLM Provider — choose from Anthropic, OpenAI, Google, or X
- Model — select the specific model (e.g., Claude Sonnet, GPT-4o, Gemini Pro)
- Add capability nodes from the palette as needed
- Changes are auto-saved — no save button needed
[Screenshot: Core node configuration panel showing name, provider, and model selection]
Node Categories
Core Node
Every character has one core node — Anima for AI characters or Naima for user entities. Naima is short for "No Anima" — it represents you in the system rather than an AI character, so it has no LLM fields since it doesn't generate responses. This core node is always present and cannot be removed.
The core node holds the character's name, LLM provider, model selection, and basic identity.
Attribute Nodes (Square Shape)
| Node | Purpose |
|---|---|
| Definition | System prompts, personality description, and behavioral instructions |
| Token Limits | Maximum input tokens, output tokens, and context window size |
Capability Nodes (Star Shape)
These enable specific features. Adding a node to the canvas activates that capability for the character.
| Node | What It Enables |
|---|---|
| Memory | Semantic memory — the character remembers past conversations |
| M2M | Machine-to-machine messaging with other characters |
| Autothink | Multi-step reasoning with 14 thinking strategies |
| Qonscious | Consciousness state machine (coherence, arousal, valence) |
| Self-Modify | Character can rewrite its own system prompts as it learns |
| Spark | Background sub-agent spawning for parallel tasks |
| Agentic | Multi-turn autonomous tool-using workflows |
| Qrawl | AI-native web browsing and content extraction |
| FractalMind | Recursive multi-directional thinking |
| Voice | Text-to-speech and speech-to-text |
| Qleph | Relational micro-language computation |
| Clipboard | Session-scoped named storage slots |
| Variable Store | Persistent key-value storage across sessions |
Tool Nodes
| Node | What It Enables |
|---|---|
| Terminal | Execute shell commands on the host machine |
| MCP | Access to 165+ tools via Model Context Protocol (see MCP Tools) |
MCP tools are organized into category flyouts: Dev (GitHub, Git, Playwright), Cloud (Docker, Kubernetes, AWS), Data (Redis, MongoDB, PostgreSQL), Comms (Telegram, WhatsApp, Email, Slack, Discord), Productivity (Trello, Notion, Jira), Security (1Password, Vault), Search (DuckDuckGo, Brave), and more.
Cortex Nodes (Diamond and Octagon Shapes)
Memory consolidation processors and modes appear in flyout menus under the Cognitive Design tab. They require the Memory node to be active.
- Processors (9 types) — define how memories are processed (decay, compression, pattern clustering, deduplication, etc.)
- Modes (8 types) — define processing styles (meditation, psychedelic, metalhead, contemplation, supermind, hyperfocus, nostalgia, archive)
See Cortex Modes for details on each mode.
Dependencies
Some nodes require other nodes to be present:
| Node | Requires | What Happens If Missing |
|---|---|---|
| Cortex Processors | Memory | Greyed out in palette with "REQUIRES MEMORY" label |
| Cortex Modes | Memory | Greyed out in palette with "REQUIRES MEMORY" label |
| Perception Triggers | Thalamus | Greyed out until Thalamus node is added |
When you add a required dependency (e.g., Memory), dependent nodes automatically become available — no manual linking needed.
The Sequencer
The right panel of the builder is the Sequencer — where you define automated tasks for the character.
Each task has:
- Trigger — what starts the task (timer/cron schedule, webhook, M2M message, or manual)
- Steps — sequential actions (LLM prompts, tool invocations, waits, conditions)
- Execution Limits — maximum steps, token budget, and timeout settings
Tools available in the Sequencer come from nodes on the canvas. If a capability isn't on the canvas, it won't appear in the step dropdown.
Key principle: Canvas defines WHO the character IS. Sequencer defines WHAT the character DOES automatically.
Tips
- Start simple. Create a character with just the core node and Definition. Add capabilities one at a time as you need them.
- Nodes are toggles. Adding a node enables a feature; removing it disables it. The character's behavior changes immediately.
- Position is cosmetic. Where you drag nodes on the canvas doesn't affect functionality — it's purely for visual organization.
- Auto-save is always on. Every change is saved immediately. There's no "discard changes" — if you make a mistake, edit it back.
