WASPIDWASPID

Documentation

Set up WASPID

WASPID is the Enterprise AI Workforce OS — one backend across a Web dashboard, a VS Code editor-tab webview, and a terminal CLI, wrapped in self-healing Loop Engineering.

Overview

Create agents, run them step-by-step, and let the Loop Engineer (Claude/your provider) diagnose failures and propose a fix you approve from email, the web, VS Code, or the terminal. Agents are versioned records, every execution is traceable, costed, and recoverable.

Install the CLI

The CLI ships as the global npm package waspid.

# Global install (use a user-writable npm prefix if needed)
npm install -g waspid

# Or run from the repo without installing:
npm run compile --workspace waspid
node packages/cli/dist/index.js --help

Connect a provider

waspid /model walks you through provider → model → API key. The key is validated, AES-256-GCM encrypted, and stored in ~/.waspid/config.json. Providers: NVIDIA, Anthropic, OpenAI, Google, Groq, Mistral, xAI, Moonshot (Kimi), DeepSeek.

waspid /model
# 1) pick a provider  →  2) pick a model  →  3) paste your API key
# ✓ Connected to NVIDIA — nvidia-nemotron-nano-9b-v2

Providers & models

WASPID connects to any of these providers. Connect a key once (CLI /model or the web Integrations page) and every agent runs on it — across the web dashboard, the VS Code webview, and the terminal.

NVIDIA

4 models
nvidia-nemotron-nano-9b-v2nvidia/llama-3.1-nemotron-70b-instructmeta/llama-3.3-70b-instructdeepseek-ai/deepseek-r1

API key: build.nvidia.com

Anthropic

4 models
claude-opus-4-8claude-opus-4-7claude-sonnet-4-6claude-haiku-4-5

API key: console.anthropic.com

OpenAI

4 models
gpt-5.5gpt-5.4gpt-5.4-minio3

API key: platform.openai.com/api-keys

Google

3 models
gemini-2.5-progemini-2.5-flashgemini-2.0-flash

API key: aistudio.google.com/apikey

Groq

4 models
llama-3.3-70b-versatilellama-3.1-8b-instantmixtral-8x7b-32768gemma2-9b-it

API key: console.groq.com/keys

Mistral

3 models
mistral-large-latestmistral-small-latestcodestral-latest

API key: console.mistral.ai/api-keys

xAI (Grok)

3 models
grok-3grok-3-minigrok-2-vision

API key: console.x.ai

Moonshot AI (Kimi)

2 models
kimi-k2moonshot-v1-128k

API key: platform.moonshot.cn

DeepSeek

2 models
deepseek-chatdeepseek-reasoner

API key: platform.deepseek.com

Terminal REPL

After connecting, choose Terminal to open the REPL — chat grounded in your real file tree, and create/run agents.

> what files are in this directory      # grounded chat (reads disk)
/agents     list / create / manage agents
/run <id>   give an agent a task (streams output)
/status     provider, model, agent count
/help       all commands     /exit  quit

Agent creation asks only name, system prompt, and max steps. When a task fails, the Loop Engineer proposes a fix you can approve and retry.

Web & IDE

The web dashboard and the VS Code extension render the same React app. In VS Code, set waspid.baseUrl and open the Waspid panel from the activity bar — the dashboard loads in an editor tab (?source=vscode drops the browser chrome).

Environment

Server env (set in .env locally, and in Vercel for hosting):

DATABASE_URL / SUPABASE_URL / SUPABASE_ANON_KEY / SUPABASE_SERVICE_KEY
NEXT_PUBLIC_SUPABASE_URL / NEXT_PUBLIC_SUPABASE_ANON_KEY   # Supabase Auth (Google)
OPENAI_API_KEY / ANTHROPIC_API_KEY
RESEND_API_KEY / TWILIO_* / RAZORPAY_*
ENCRYPTION_KEY (openssl rand -hex 32) / NEXT_PUBLIC_URL / WASPID_ENV

API reference

REST endpoints (session-authenticated unless noted): /api/agents CRUD, /api/execute, /api/executions[/id], /api/loop/generate-fix, /api/approvals, /api/integrations, webhooks /api/webhooks/{approve,whatsapp,razorpay}, and a public /api/health.