pokegent

◓ Pokégent

Terminal dashboard + shareable HTML Trainer Card for your AI coding ecosystem.

Your coding agents become Pokémon species, MCP servers become TMs/HMs, installed skills fill your Move Tutor log, and token burn tracks as PP.

Features · Installation · Usage · Trainer Card · Scoring · Privacy

Pokégent scans your local machine to give you a live picture of your AI tooling landscape. It detects running AI processes, finds installed MCP servers, discovers skills, tallies model usage from shell history, and measures token burn rates — all locally with zero outbound requests.

Note
Pokégent runs 100% locally. It reads the process table, configuration directories, and log files on your machine and renders the results in your terminal. No telemetry, no network calls, no data leaves your computer.

Features

• 16 agent detectors — Mewtwo (Claude Code), Pikachu (Gemini CLI), Ditto (OpenCode), and more. Each detected process gets a Pokémon species, sprite, and live CPU/memory stats.
• MCP server discovery — Scans ~/.claude.json, ~/.cursor/mcp.json, ~/.opencode/config.json, ~/.n8n/mcp.json, and per-project MCP configs. Full arsenal displayed as TM-numbered chips.
• Skills inventory — Discovers installed skills from ~/.claude/skills, ~/.agents/skills, and ~/.config/opencode/skills.
• Model usage charts — Parses terminal history and log files for model mentions. Renders horizontal HP-style bars with relative-to-top scaling.
• Token burn analytics — Session count from .jsonl transcripts, token velocity (PP/min), input/output splits, estimated monthly cost.
• Shareable HTML Trainer Card — Self-contained card with trainer ID, full MCP arsenal grid, team roster with sprites, PP stat tiles, badges, and one-click COPY STATS + SHARE ON X.
• Scoring engine — 0–1000 points across agents, MCP, models, skills, and burn. Rarity tiers from Beginner Trainer to Mythical Champion. 7 unlockable badges.

Installation

# Run directly (zero install)
npx pokegent

# Or install globally
npm install -g pokegent

From source

git clone https://github.com/shafiqimtiaz/pokegent.git
cd pokegent
npm install
npm run build

Usage

# Start the TUI dashboard (live scan)
pokegent

# Demo mode with realistic mock data
pokegent --demo

# Export scanned data as JSON
pokegent --json

# Generate the shareable HTML Trainer Card
pokegent --share

Key	Action
q	Quit
r	Force scan refresh

Trainer Card

The --share flag generates a self-contained HTML file at ~/Desktop/pokegent-card.html.

The card includes:

• Trainer identity — Your username + ID No. derived from token burn.
• Score strip — Total score, rarity rank, per-dimension breakdown.
• Team roster — All detected agents with Pokémon sprites from PokeAPI, species name, CLI name, and state chip (active/resting/in box).
• MCP arsenal — Full TM grid (up to 18 servers) with tool counts and descriptions on hover.
• Skills — Installed skills listed as tagged chips.
• Movepool — Model usage with animated HP bars, relative scaling.
• PP Burn tiles — Tokens, cost, velocity, session count.
• Badges — Unlocked achievements with holo shine animation.
• Share strip — COPY STATS button (clipboard) + SHARE ON X (prefilled tweet).

pokegent --demo --share   # Generate card with mock data

Scoring

Dimension	Max	How
Agents	350 pts	75 pts per running agent (cap 4) + 50 if 3+ active
MCP servers	200 pts	10 pts per server (cap 15) + 1 per tool (cap 50)
Models	200 pts	30 pts per model (cap 5) + 50 for 3+ providers
Skills	100 pts	5 pts per skill (cap 20)
Token burn	250 pts	Velocity tiers + session milestones

Rarity tiers

Score	Title
900+	🌟 Mythical Champion — Top 1%
750+	💎 Shiny Legendary — Top 5%
600+	🥇 Pokémon Master — Top 15%
400+	🥈 Gym Leader — Top 35%
200+	🥉 Elite Trainer — Top 60%
<200	🌱 Beginner Trainer

Badges

Badge	Condition
🏆 Pokédex Master	10+ MCP servers
🦄 Legendary Trainer	3+ agents running
🧬 Hybrid Evolution	3+ model providers
🔥 Blast Burn	10K+ tokens/min velocity
💎 Elite Four	100+ sessions
🌐 Safari Zone Master	5+ models detected
📚 Move Tutor	20+ skills installed

What gets scanned

Agent detectors (16)

Pokémon	Platform	Process	Config paths
🔮 Mewtwo	Claude Code	claude, claude-code	~/.claude, ~/.claude.json
🍃 Venusaur	Codex	codex, openai-codex	~/.codex
🐢 Blastoise	Copilot CLI	copilot, github-copilot	~/.copilot
⚡ Pikachu	Gemini CLI	gemini, gemini-cli	~/.gemini
🦊 Eevee	Cursor	cursor	~/.cursor
🔥 Charizard	Amp	amp, amp-cli	~/.amp
🦎 Charmander	Cline	cline	~/.cline
👻 Gengar	Roo Code	roo, roo-code	~/.roo
🐻 Snorlax	Kilo Code	kilo, kilo-code	~/.kilo
🦇 Zubat	Kiro	kiro	~/.kiro
🎈 Jigglypuff	Crush	crush	~/.crush
🍮 Ditto	OpenCode	opencode	~/.opencode
💪 Machamp	Factory Droid	factory, factory-droid	~/.factory-droid
🐉 Rayquaza	Antigravity	antigravity, ag-cli	~/.antigravity
⛵ Lapras	Kimi CLI	kimi, kimi-cli	~/.kimi
🐉 Dragonite	Qwen Code	qwen, qwen-code	~/.qwen

MCP config paths

• ~/.claude.json (including per-project mcpServers)
• ~/.claude/mcpServers.json
• ~/.claude/config.json
• ~/.config/claude/settings.json
• ~/.cursor/mcp.json
• ~/.opencode/config.json
• ~/.roo/mcp.json
• ~/.kilo/mcp.json
• ~/.n8n/mcp.json

Skill scan directories

• ~/.claude/skills
• ~/.agents/skills
• ~/.config/opencode/skills

Model patterns (24+)

Anthropic: Claude Fable 5, Sonnet 5, Opus 4.8/4.5/4.1, Sonnet 4.5, Haiku 4.5, 4 Opus, 4 Sonnet, 3.7 Sonnet, 3.5 Sonnet, 3 Haiku.
OpenAI: GPT-5, GPT-4.1, GPT-4o, o4-mini, o3, o3-mini.
Google: Gemini 3, 2.5 Pro, 2.5 Flash.
Other: DeepSeek V3/R1, Qwen 3, Llama 4.

Privacy

Pokégent runs 100% locally. No data leaves your machine. No telemetry, no analytics, no outbound network requests during scanning. All process table reads, configuration file scans, and log parsing happen entirely on your computer. The --share card is written as a local HTML file on your desktop — sharing it is your choice.