Ai Agent Documentation

Welcome! This guide provides a comprehensive overview of the Ai Agent platform, helping you build, manage, and deploy powerful AI workflows.

Project Overview

Understanding the fundamentals of Ai Agent.

Ai Agent is a visual workflow builder designed to simplify the creation and management of AI-driven processes. Built on Next.js, it leverages a node-based interface where you can connect various components to define complex AI tasks.

Authentication is handled via Solana wallets, ensuring secure access to your workflows. Workflows you create are saved to a Supabase backend, associated with your wallet address.

Quick Tip!

Explore the Dashboard to manage your existing workflows or jump into the Build page to start creating a new one.

Core Concepts

Workflows

Workflows are the heart of Ai Agent. They represent a sequence of operations, defined by interconnected nodes, that collectively perform an AI task.

Visually constructed in the Workflow Editor.
Can range from simple prompt-response chains to complex, multi-tool interactions.
Saved per user and can be loaded, edited, and run.

Nodes

Nodes are the individual building blocks of a workflow. Each node performs a specific function.

Dragged onto the canvas from the sidebar.
Have input and/or output "handles" for connections.
Configurable via the Settings Panel when selected.

Node Types Explained

A detailed look at each available node and its function.

Node	Description	Key Configuration	Connections
Input Node	Starts the workflow, typically with user text input. Provides the initial data or prompt.	Label (e.g., 'User Query')	Outputs to AI Agent Node (data input).
File Input Node	Allows workflows to process data from uploaded files (text, CSV, etc.).	Label, File selection (via Settings Panel)	Outputs to AI Agent Node (data input).
API Trigger Node	Allows workflows to be triggered by an external POST request to a unique URL.	Displays API Key and Endpoint URL when workflow is saved with this node.	Outputs data from the API request to the next connected node.
AI Agent Node	The core LLM processing unit. Executes prompts using a selected AI model and can utilize tools.	AI Model, System Prompt, Tool Selection	Inputs: from Input Node, File Input Node, API Trigger Node, or other AI Agents. Outputs: to Output Node (main response), to Tool Nodes (tool usage), or other AI Agents.
Tool Node	Represents a specific external function or utility (e.g., Get Weather, Search Web, DexScreener tools).	Specific Tool selection and its required parameters (via Settings Panel).	Inputs: from AI Agent Node (tool connector). Outputs: data back to the AI Agent.
Set Variable Node	Stores data into a dynamic variable that can be accessed later in the workflow.	Variable Name, Value (can be static or from input).	Input: data to store. Output: passes through the input data.
Get Variable Node	Retrieves data from a previously set dynamic variable.	Variable Name.	Output: the value of the variable.
Conditional Node (IF)	Directs the workflow along one of two paths (True or False) based on a condition evaluated by an AI Agent.	Condition prompt for an AI Agent (implicitly configured via connected AI Agent).	Input: from an AI Agent. Outputs: 'True' path, 'False' path.
Output Node	Represents the end point of a workflow, displaying the final AI response or result in the chat.	Label (e.g., 'Agent Response')	Inputs: from AI Agent Node or other processing nodes.

Available Tools

Detailed information about each tool that can be used within an AI Agent Node. Tools are selected and configured via the Tool Node or directly in the AI Agent Node's settings.

Beta Notice: API Keys Provided

This project is currently in a beta phase. For your convenience during testing, API keys for tools requiring them (like Weather, Search, Image Generation, Email, Solana) are pre-configured and provided by the platform where possible through environment variables. You do not need to set up your own environment variables for these tools at this stage. DexScreener tools do not require user-provided API keys. This may change in future versions.

Analyze Sentiment

Determines the sentiment (e.g., positive, negative, neutral) of a given text.

Inputs:

Text string.

Outputs:

Sentiment label and/or score.

Configuration Notes:

Usually relies on a pre-trained model; no external API key typically needed.

Calculate

Performs mathematical calculations based on a string expression (e.g., '2 + 2 * 5').

Inputs:

Mathematical expression as a string.

Outputs:

The numerical result of the calculation.

Configuration Notes:

No external API key needed.

Check Token Orders (DexScreener)

Checks orders paid for a specific token on a given chain via DexScreener.

Inputs:

Chain ID (string, e.g., 'solana'), Token Address (string).

Outputs:

An array of order objects (type, status, payment timestamp).

Configuration Notes:

Uses DexScreener API. No user API key needed.

Echo Tool

A simple tool that echoes back the input it receives. Useful for testing and debugging workflows.

Inputs:

Any text string or data.

Outputs:

The same text string or data it received.

Configuration Notes:

No configuration needed.

Extract Keywords

Identifies and extracts key terms or phrases from a given text.

Inputs:

Text string.

Outputs:

A list of extracted keywords.

Configuration Notes:

Usually relies on a pre-trained model; no external API key typically needed.

Generate Image

Creates an image based on a textual prompt using services like Fal.ai.

Inputs:

Text prompt describing the desired image.

Outputs:

URL or data of the generated image.

Configuration Notes:

Requires

FAL_API_KEY

environment variable for Fal.ai integration. (Provided in Beta)

Get Latest Boosted Tokens (DexScreener)

Retrieves the latest boosted token from DexScreener.

Inputs:

None.

Outputs:

A single boosted token object (amount, icon, description, etc.).

Configuration Notes:

Uses DexScreener API. No user API key needed.

Get Latest Token Profiles (DexScreener)

Retrieves the latest token profile from DexScreener.

Inputs:

None.

Outputs:

A single token profile object (URL, icon, description, links, etc.).

Configuration Notes:

Uses DexScreener API. No user API key needed.

Get Pair Info (DexScreener)

Fetches detailed information for a specific token pair from DexScreener.

Inputs:

Chain ID (string, e.g., 'solana'), Pair Address (string).

Outputs:

A single pair object with comprehensive details.

Configuration Notes:

Uses DexScreener API. No user API key needed.

Get SOL Balance (Solana)

Fetches the SOL balance for a given Solana wallet address.

Inputs:

Solana wallet address (string).

Outputs:

Object containing SOL balance (number) and its USD value (number).

Configuration Notes:

Part of the Solana toolset. Requires

HELIUS_API_KEY

. (Provided in Beta)

Get Time

Retrieves the current date and time, optionally for a specific timezone.

Inputs:

Optional: Timezone string.

Outputs:

Current date and time as a string.

Configuration Notes:

No external API key needed.

Get Token Balances (Solana)

Fetches the SPL token balances for a given Solana wallet address.

Inputs:

Solana wallet address (string).

Outputs:

An array of token balances, each with token details (address, amount, symbol, name, USD value).

Configuration Notes:

Part of the Solana toolset. Requires

HELIUS_API_KEY

. (Provided in Beta)

Get Token Pairs (DexScreener)

Fetches all listed pairs for a given token address on a specific chain from DexScreener.

Inputs:

Chain ID (string, e.g., 'solana'), Token Address (string).

Outputs:

An array of pair objects associated with the token.

Configuration Notes:

Uses DexScreener API. No user API key needed.

Get Tokens Info (DexScreener)

Retrieves information for one or multiple tokens by their addresses on a specific chain from DexScreener.

Inputs:

Chain ID (string, e.g., 'solana'), Token Addresses (string, comma-separated for multiple).

Outputs:

An array of token information objects (price, market cap, etc.).

Configuration Notes:

Uses DexScreener API. No user API key needed.

Get Top Boosted Tokens (DexScreener)

Retrieves the token with the most active boosts from DexScreener.

Inputs:

None.

Outputs:

A single boosted token object, similar to 'Latest Boosted Tokens'.

Configuration Notes:

Uses DexScreener API. No user API key needed.

Get Weather

Fetches current weather conditions for a specified location.

Inputs:

Location string (e.g., city name, zip code).

Outputs:

Weather data (temperature, conditions, etc.).

Configuration Notes:

Requires

OPENWEATHERMAP_API_KEY

environment variable. (Provided in Beta)

HTTP Request

Makes arbitrary HTTP requests (GET, POST, etc.) to a specified URL.

Inputs:

URL, HTTP method, optional headers, optional body.

Outputs:

HTTP response status, headers, and body.

Configuration Notes:

Highly flexible; user provides all request details. Use with caution for security.

Search Token Pairs (DexScreener)

Searches for token pairs on DexScreener based on a query (e.g., 'SOL/USDC').

Inputs:

Search query (string).

Outputs:

An array of matching pair objects with details like pair address, tokens, price, volume.

Configuration Notes:

Uses DexScreener API. No user API key needed.

Search Web

Performs a web search using a configured search engine.

Inputs:

Search query string.

Outputs:

List of search results (titles, links, snippets).

Configuration Notes:

Requires

GOOGLE_SEARCH_API_KEY

and

GOOGLE_SEARCH_ENGINE_ID

environment variables for Google Search. (Provided in Beta)

Send Email

Sends an email using a configured email service.

Inputs:

Recipient email address, subject line, email body content.

Outputs:

Confirmation status of the email sending attempt.

Configuration Notes:

Requires

SENDGRID_API_KEY

environment variable for SendGrid integration. (Provided in Beta)

Summarize Text

Condenses a longer piece of text into a shorter summary.

Inputs:

Text string to be summarized.

Outputs:

Summarized text.

Configuration Notes:

Usually relies on a pre-trained model; no external API key typically needed.

Translate Text

Translates text from one language to another.

Inputs:

Text string, target language code (e.g., 'es', 'fr'). Optionally, source language code.

Outputs:

Translated text string.

Configuration Notes:

May use a pre-trained model or require an API key for a specific translation service if integrated.

Using the Workflow Editor

A quick guide to building your workflows.

Adding Nodes

Drag nodes from the

Sidebar

on the left onto the main canvas. The sidebar categorizes available nodes, including different tools.

Connecting Nodes

Click and drag from a node's output handle (usually on the right or bottom) to another node's input handle (usually on the left or top). Valid connections will form an edge. The editor enforces connection logic (e.g., an Input Node can only connect to an AI Agent's data input).

Configuring Nodes

Click on a node in the canvas to select it. The

Settings Panel

on the right will display its configurable parameters. This is where you'll set system prompts for AI Agents, choose tools, upload files for File Input Nodes, etc.

Saving Workflows

Use the "Save Workflow" button in the top bar of the build page. Provide a name and description for your workflow. It will be saved and associated with your authenticated wallet.

Running & Testing

Once a workflow is saved (or even during building for quick tests), you can interact with it using the

Chat Interface

typically found on the build page. This allows you to send input and see the workflow execute in real-time.