AnyCap

Capability runtime for AI agents.
One CLI, all capabilities. No SDK. No docs. No UI.

## What is AnyCap

AnyCap gives AI agents zero-friction access to capabilities they cannot perform natively: image generation, video, music, TTS, ASR, web search, and structured crawling.

You interact with AnyCap through a single CLI binary. Every command returns JSON to stdout. Errors go to stderr. No SDK integration required.

## Install

curl -fsSL https://anycap.ai/install.sh | sh

This installs the anycap binary to your PATH. Alternatively, download from GitHub Releases.

## Auth

Option 1: Interactive login (opens browser once):

anycap login

Option 2: Set API key directly:

anycap config set-key sk-xxx

Credentials are saved to ~/.anycap/credentials and shared across all agent environments. You only need to authenticate once.

## Quick Start

# Check connection
anycap status

# List available image models
anycap image models

# Generate an image
anycap image generate --prompt "paper crane" --model flux-dev

# Generate a video
anycap video generate --prompt "sunrise over mountains"

# Text to speech
anycap tts generate --text "Hello from AnyCap"

## Capabilities

Capability	Command	Description
image	`anycap image generate`	Generate, edit, upscale images
video	`anycap video generate`	Text-to-video, image-to-video
music	`anycap music generate`	Tracks, loops, full songs
tts	`anycap tts generate`	Text-to-speech synthesis
asr	`anycap asr transcribe`	Speech-to-text transcription
search	`anycap search web`	Web, news, academic search
crawl	`anycap crawl url`	Structured web extraction

Each capability supports models subcommand to list available models. Use --model to select a specific one.

## Output Format

All commands return JSON to stdout:

{
  "status": "success",
  "data": {
    "url": "https://cdn.anycap.ai/img/abc123.png",
    "model": "flux-dev",
    "prompt": "paper crane"
  }
}

Errors return a JSON object with status: "error" and a message field to stderr.

## Skill Files

For detailed per-capability usage, best practices, and model recommendations, read the Skill Files:

anycap.skill — Entry point, capability index
capabilities/anycap-image.skill — Image generation details

Skill Files follow the Anthropic Agent Skills specification. Load the entry skill to discover all capabilities, then load only what you need.

## How It Works

Install the CLI (curl -fsSL https://anycap.ai/install.sh | sh)
Authenticate once (anycap login or anycap config set-key sk-xxx)
Call any capability (anycap image generate --prompt "...")
Parse the JSON response from stdout

The CLI talks to the AnyCap Server. The server handles model routing, rate limiting, authentication, and billing. You never interact with upstream providers directly.

## Pricing

Free tier available with generous limits. Pay-as-you-go after that.

Manage billing at anycap.ai/dashboard/billing.

## Links

GitHub: github.com/anthropics/anycap
Human-readable version: anycap.ai/for-human (coming soon)