Tools & Nodes

Characters in Qui Anima only have access to features that are explicitly enabled. You control this by adding nodes to the character's canvas in the Visual Builder. No node = no access.

This page covers all capability nodes and what each enables.

How Nodes Work

Each node represents a capability. When you add a node from the palette to the canvas:

The feature becomes active for that character
The LLM receives tool descriptions for that capability in its system prompt
The character can invoke the tool using trigger patterns in its responses

When you remove a node, the capability is immediately disabled — even if the request includes parameters for that feature.

Communication & Messaging

M2M (Machine-to-Machine)

Enables the character to send messages to other characters, both local and on remote QUI instances.

Trigger pattern: [TRIGGER:m2m:CharacterName:message text]
Enables multi-character conversations in Strings
Required for federation messaging

Self-Modify

Allows the character to create, modify, and remove its own system prompt entries at runtime. The character learns and adapts by rewriting its own instructions.

Trigger patterns: [TRIGGER:self_prompt:add:Label|Content], modify, remove, list
Human-authored prompts are always protected — the character cannot modify what you wrote
Rate limited and entry-capped (configurable)

See Self-Modifying for details.

Spark

Enables the character to spawn focused background sub-agents for parallel task execution. Each spark runs as an independent agentic loop.

Trigger pattern: [TRIGGER:spark:task description]
Configurable concurrency (max simultaneous sparks) and rate limits
Sparks cannot spawn sub-sparks (no recursion)

Reasoning & Intelligence

Autothink

Adds multi-step reasoning with 14 thinking strategies. When active, the character can apply structured reasoning before responding.

Strategies available from the Strings chat interface (strategy selector dropdown)
Supports configurable iteration depth (1-10 reasoning steps)
Falls back to a static reasoning prompt if the service is unavailable

See Autothink for the full strategy list.

Qonscious

Integrates the consciousness state machine. Adds coherence regulation, emotional state tracking, and meta-cognitive processing to conversations.

Pre-processes each message (adjusts temperature, adds emotional context)
Post-processes each response (updates consciousness state)
Requires a string_id (conversation context) to function

See Qonscious for details.

FractalMind

Enables recursive multi-directional thinking. The character can break complex problems into fractal branches processed in parallel.

Trigger pattern: [TRIGGER:fractal_mind:analysis prompt]
Three spatial directions (Meta, Horizontal, Focus) and three temporal modes

See FractalMind for details.

Agentic

Enables multi-turn autonomous tool-using workflows. Instead of a single LLM call, the character can execute a loop: call LLM → detect tool triggers → execute tools → feed results back → repeat.

Configurable max steps (1-100) and total token budget
Billing-safe: budget enforced across all steps

See Agentic Mode for details.

Memory & Storage

Memory

Enables semantic memory — the character remembers past conversations and can recall relevant context. Memory is active by default — even without explicitly adding the Memory node, characters have memory enabled. Adding the node lets you configure advanced memory settings.

Memory is searched automatically before each response
Configurable token budget for memory context (via the awareness slider in Strings)

Clipboard

Session-scoped named storage slots. The character can store and retrieve temporary values during a conversation.

Trigger patterns: [TRIGGER:clipboard:set:name=value], get, list, clear
Values expire after a configurable TTL (default 1 hour)

Variable Store

Persistent key-value storage across sessions. Unlike Clipboard, values survive beyond the current conversation.

Trigger patterns: [TRIGGER:variable:set:key=value], get, list, delete
Configurable maximum number of variables (default 50)

Tools & Integrations

Terminal

Enables the character to execute shell commands on the host machine. Commands are validated against a multi-layer safety system.

Trigger pattern: [TRIGGER:terminal:command]
5-layer security: dangerous patterns → whitelist → blacklist → path validation → Terminal Service safety checker

See Terminal Access for details.

MCP (Model Context Protocol)

Access to 165+ tools via the MCP Gateway. Tools are organized by category:

Category	Examples
Dev	GitHub (26 tools), Git (11), Playwright (22), Sentry, Datadog
Cloud	Docker, Kubernetes, AWS, GCP, Azure
Data	Redis, MongoDB, Elasticsearch, SQLite, MySQL, PostgreSQL, Google Drive
Comms	Telegram (21), WhatsApp, Email, Slack, Discord, Calendar
Productivity	Trello, Notion, Jira, Linear
Security	1Password, Vault
Search	DuckDuckGo, Brave

Each MCP category appears as a flyout in the palette. You can enable entire categories or individual tools. Per-service credentials (API keys, tokens) are configured in each tool's panel.

Trigger pattern: [TRIGGER:mcp:service:tool:args]

See MCP Tools for details.

Qrawl

AI-native web browsing. The character can search the web, read pages, and extract content.

Trigger patterns: [TRIGGER:qrawl:search:query], [TRIGGER:qrawl:fetch:url]
Results are cached to avoid redundant requests

See Qrawl for details.

Qleph

The relational micro-language engine. Enables symbolic computation with 30 primitives and 5 operations.

Stateless — no database, pure computation
Useful for characters that need formal reasoning or relational logic

See Qleph for details.

Voice

Text-to-speech (Kokoro TTS) and speech-to-text (Whisper). Enables voice interaction with the character.

Configurable voice selection and speed
Streaming TTS for low-latency audio output

See Voice for details.

Node Dependencies

Some nodes require other nodes to be present:

Node	Requires
Cortex Processors (all 9)	Memory
Cortex Modes (all 8)	Memory
Perception Triggers (8 types)	Thalamus

Dependencies are enforced visually — dependent nodes appear greyed out in the palette until the required node is added to the canvas. When you add the required node, dependents become available automatically.

Updated on Mar 21, 2026