AgentUI

MAIN

TASK CENTER

Task Center

Welcome to AgentUI — an AI interface with specialized agents.

New session:

Or continue:

The assistant can automatically open specialized agents for different tasks:

BASE — Basic tasks with search
CODE — Programming and data analysis
RESEARCH — Deep research and information gathering
IMAGE — Image generation and editing
COMPUTER (coming soon) — Computer control and automation

Agent

Main agent

User query

Routing to sub-agents...

Research

Sub-agent

12 turns

search"topic overview"

read4 sources

Synthesizing...

result

Code

Sub-agent

4 turns

codeimport numpy as np

codeplt.show()

result2 figures

Image

Sub-agent

2 turns

Generating image...

result1 image

Summarize results

When an agent is opened, you'll see a widget you can click to jump to it. A pulsing dot on the tab indicates active generation.

SETTINGS

Agent Models

Models & Providers

PROVIDERS API endpoints for LLM services

MODELS Models available from your providers

API Keys

E2B API KEY Required for code execution in CODE agents

SERPER API KEY Required for web search in RESEARCH agents

HUGGINGFACE TOKEN Required for image generation in IMAGE agents

Agent Settings

IMAGE GENERATION MODEL Model for text-to-image generation

IMAGE EDIT MODEL Model for image-to-image editing

RESEARCH SUB-AGENT MODEL Smaller/faster model for analyzing web pages during research

RESEARCH PARALLEL WORKERS Number of web pages to analyze in parallel (default: 8)

RESEARCH MAX WEBSITES Max websites to analyze per session (default: 50)

Theme

Forest

Sapphire

Ocean

Midnight

Steel

Depths

Ember

Noir

Eclipse

Terminal

DEBUG: Message History

No message history available yet.

FILES

Hidden

Loading...

Click: insert path Double-click: expand Drag: to input

SESSIONS

Current session:

New session:

Switch to:

Loading...

NAME

ENDPOINT

API TOKEN

DISPLAY NAME

PROVIDER

MODEL ID The exact model identifier used in API calls

EXTRA PARAMETERS Optional JSON object with extra parameters for API calls (e.g., enable_thinking, temperature)

Supports vision / image input