GEO Agent Workspace

An Agentic Workspace for tracking your brand's visibility across AI-powered search engines.

• Query AI platforms (ChatGPT, Claude, Perplexity, Gemini) with real web search to see if and how they mention your brand vs competitors
• Track visibility, share-of-voice, sentiment, and recommendation rate across TOFU/MOFU/BOFU queries
• Auto-research competitors via web scraping to find content gaps and generate the exact queries where you're losing visibility

Prerequisites

• Bun runtime
• At least one AI platform API key (Anthropic required, others optional)

Quick Start

git clone <repo-url> && cd geo-agent
cd agentic-geo-dashboard
cp .env.example .env   # Add your API keys
bun install
bun start

API Keys

Add these to agentic-geo-dashboard/.env:

Key	Required	Platform
ANTHROPIC_API_KEY	Yes	Claude (also used for LLM-based mention extraction)
OPENAI_API_KEY	No	ChatGPT
PERPLEXITY_API_KEY	No	Perplexity
GEMINI_API_KEY	No	Gemini
COMPOSIO_API_KEY	No	Web search and scraping for competitor research

Usage

The dashboard is a terminal UI built with OpenTUI. Run from the agentic-geo-dashboard/ directory:

bun start              # Launch the dashboard
bun dev                # Launch with watch mode (auto-reload on changes)

Screens

Navigate between screens with number keys 1-5:

Key	Screen	Description
1	Overview	Summary of latest simulation: visibility scores, mention rates, platform breakdown
2	Competitors	Run competitor research via Composio web search. Shows competitor gaps and opportunities
3	Simulate	Run a new visibility simulation across all configured platforms and queries
4	Results	Detailed results from the latest simulation: per-query breakdown, share-of-voice, positions
5	History	Browse past simulation runs and track trends over time

Keyboard Shortcuts

• 1-5 — Switch screens
• q — Quit
• Ctrl+E — Edit configuration (brand, website, competitors, queries)
• Ctrl+S — Save (in Setup screen)

Headless Commands

bun run simulate    # Run simulation without the TUI
bun run research    # Run competitor research and save results
bun run schedule    # Start the scheduler (runs simulations on an interval)

Composio

The competitor research agent uses Composio for web search and scraping — it discovers competitor content, analyzes their online presence, and identifies content gaps without needing custom scraping code.

Composio is configured as an MCP server in .claude/settings.json. To set it up:

1. Get your API key from the Composio Dashboard
2. Add it to .env as COMPOSIO_API_KEY
3. When the agent needs a tool, Composio handles authentication inline — no pre-configuration needed

See the Composio docs for the full list of supported apps and toolkits.