Node Reference

ThinkThing has over 143 node types organized into categories. This reference covers each category and its key nodes. Nodes marked with (requires Anima) need an Anima character node connected to function.

Anima

Node	Purpose
Anima	AI Character connection — links the graph to a specific character for all LLM-powered nodes. Optionally enables agentic mode for autonomous multi-step tool use within individual node executions.

Every graph that uses cognition, terminal, or MCP nodes needs at least one Anima node.

Control Flow

These nodes manage the structure and flow of your graph. No LLM calls — pure logic.

Node	Purpose
Start	Entry point. Every graph begins here.
Timer Start	Delayed start with debounce/rate limiting.
Schedule Start	Cron-based scheduler — run graphs on a schedule.
Webhook	HTTP trigger — start a graph via an external POST request.
End	Termination point. Content reaching End is the graph's output.
Hub	Fan-in (merge multiple paths) or fan-out (distribute to multiple paths).
Control	Human checkpoint — pauses execution for approve, revise, or reject.
Fallback	Entry point for revised content when a Control node rejects.
Combine	Concatenate multiple inputs into a single output.
Split	Break input into parts using a delimiter.
Buffer	Collect N items before releasing them all at once.
Delay	Hold input for a specified number of seconds.

Cognition (requires Anima)

Standard LLM-powered processing nodes. Each sends content to the connected character's LLM and returns the result.

Node	Purpose
Gate	Binary yes/no decision — routes content to one of two paths.
Choice	Multi-path routing — LLM selects from multiple options.
Loop	Loop end — LLM decides whether to continue iterating or exit.
LoopStart	Loop entry point (pairs with Loop node). Does not require Anima.
Prompt	Custom LLM prompt — the most flexible node. You define the task.
Summarize	Combine and condense multiple inputs.
Compare	Find similarities and differences between inputs.
Extract	Pull specific information from content (names, dates, facts).
Classify	Categorize content into predefined labels.
Rewrite	Rephrase content in a different style or format.
Joke	Generate humor based on input content.
Translate	Language translation.

Advanced Cognition (requires Anima)

Higher-level reasoning nodes that apply structured thinking to content.

Node	Purpose
Autothink	Apply one of 14 thinking strategies to analyze content.
Decision	Weighted decision making — evaluate options against criteria.
Evaluate	Score content against specified criteria (grading, quality assessment).
Reason	Self-prompting logical reasoning — the LLM builds its own chain of thought.
Inspector	Multi-perspective analysis — examines content from different viewpoints.
Brainstorm	Generate multiple ideas iteratively, building on previous suggestions.
Plan	Create detailed step-by-step plans from a goal.
Critic	Critical analysis — identify weaknesses, suggest improvements.

Super Cognition (requires Anima)

Specialized cognitive nodes that connect to advanced services.

Node	Purpose
FractalMind	Recursive multi-directional thinking via the FractalMind service.
Qleph Dictionary	Fetch Qleph relational micro-language dictionary for LLM context.
Qleph Engine	Validate, evaluate, and invert Qleph expressions.

Terminal (24 types, requires Anima)

Execute shell commands on the host machine. The LLM decides what commands to run based on your task description, executes them through an agentic loop, and analyzes the results.

Key nodes:

Node	Purpose
User Terminal	Direct command execution — no LLM, you type the command.
Agent Terminal	LLM-driven terminal — describe a task, the character executes commands.
Terminal Approval	Human-in-the-loop gate for command review before execution.

21 specialized terminal nodes for specific domains: Git, Python, Node.js, Rust, Go, Make, Java, System Info, File Operations, Text Processing, Processes, Network, SSH, Docker, Kubernetes, Podman, PostgreSQL, MySQL, Redis, SQLite, MongoDB.

Each specialized node scopes the LLM's system prompt to that domain (e.g., "This is a git terminal" or "This is a PostgreSQL terminal") for more focused command generation.

Memory

Node	Purpose
Memory	Store content to long-term semantic memory.
Variables	Working memory — set and get variables within the current execution.
Knowledge	Document repository — content is injected into the Anima character's context.
Recall	Retrieve memories by semantic search.
Learn	LLM modifies its own persistent instructions (self-modify via agentic loop).

Perception

Nodes that monitor external conditions and trigger workflows.

Node	Purpose
File Watcher	Watch file/directory changes.
URL Monitor	Poll a URL for content changes.
API Poller	Poll an API endpoint for status changes.
System Monitor	CPU/RAM/disk threshold alerts.
Observer	General observation node.
QRawl Search/Skim/Read/Focus	Web search and content extraction via Qrawl.

Output

Node	Purpose
Display	Show content in the execution monitor output.
File Writer	Write content to a file on disk.

Integration

Node	Purpose
Agent (M2M)	Send messages to other characters via M2M messaging.
API	Make HTTP requests to external APIs.
Speak	Text-to-speech output via Voice Service.
Listen	Speech-to-text input via Voice Service.

MCP Service Nodes (37 types, requires Anima)

Dedicated nodes for specific MCP tools, organized by category:

Category	Services
Dev	GitHub, Git, Playwright, Filesystem, Fetch, Sentry, Datadog
Cloud	Docker, Kubernetes, AWS, GCP, Azure
Data	Redis, MongoDB, Elasticsearch, SQLite, MySQL, PostgreSQL, Google Drive
Comms	Telegram, WhatsApp, Email, Slack, Discord, Calendar
Productivity	Trello, Notion, Jira, Linear
Security	1Password, Vault
Cognition	Sequential Thinking, Memory Knowledge Graph
Search	DuckDuckGo, Brave
Utility	Time

Each MCP node is pre-configured with the service's available tools. Add the node, configure credentials, and connect to an Anima character.

Multi-Agent

Node	Purpose
Broadcast	Send content to multiple characters simultaneously.
Collect	Gather responses from multiple characters.
Consensus	Analyze collected responses and find agreement or disagreement.
Delegate	Assign a specific task to a specific character.

Qonscious (14 nodes)

Consciousness state manipulation within workflows.

LLM-powered (analyze content): Q Gate, Q Watcher, Q Insight, Q Observer

Data-driven (no LLM): Q Profiler, Q Compress, Q Coherence

State mutation tools: Q Inject, Q Decay, Q Emote, Q Learn, Q Shift, Q Reset, Q Watcher Reset

These nodes interact with the Qonscious service to read, modify, and respond to a character's consciousness state during workflow execution.

Updated on Mar 21, 2026