AudioStack AdCP Agent

Connect

Add this to your Claude Desktop or Claude Code .mcp.json:

{
  "mcpServers": {
    "audiostack-adcp": {
      "url": "https://adcp.lab.maikla.com/mcp"
    }
  }
}

Or connect any MCP-compatible client to the endpoint above.

Quick start

Once connected, just ask Claude naturally:

"Create a 30-second audio ad for Nike promoting their spring running collection.
Energetic tone, targeting fitness enthusiasts. CTA: visit nike.com"

Behind the scenes, Claude calls build_creative with your brief and returns a produced audio ad with a playable URL.

Step by step

# 1. Discover what formats are available
→ list_creative_formats

# 2. Preview a script before producing
→ preview_creative { request_type: "single", message: "..." }

# 3. Produce the audio ad
→ build_creative { message: "...", brand: { domain: "nike.com" },
    target_format_id: { format_name: "audio-30s-mp3" } }

# 4. Validate against delivery specs
→ sync_creatives { account: {...}, creatives: [...] }

# 5. Check delivery metrics
→ get_creative_delivery { creative_ids: ["..."] }

Available tools

get_adcp_capabilities

Discover supported tools, formats, and channels

sync

list_creative_formats

Browse 10 audio formats (10s-60s, MP3/WAV, streaming/radio)

sync

build_creative

Generate a produced audio ad from a natural language brief

async

preview_creative

Generate a script and production plan without producing audio

sync

sync_creatives

Validate and distribute finished creatives to delivery platforms

async

get_creative_delivery

Retrieve delivery status and performance metrics

async

Endpoints

POST /mcp MCP protocol (tool calls)

GET /mcp SSE stream (session)

GET /health Health check

GET /.well-known/adagents.json AdCP agent discovery

Supported formats

All formats are 44.1kHz, mono/stereo audio:

Duration	MP3 (320kbps)	WAV	Radio
10s	✓
15s	✓	✓
20s	✓
30s	✓	✓	✓
40s	✓
60s	✓	✓

How it works

This agent implements the Ad Context Protocol (AdCP) creative workflow over MCP.

You send a natural language brief via build_creative
AudioStack's AI generates a script tailored to your brand, audience, and tone
A voice is selected and the script is synthesized with professional TTS
Background music is matched and mixed with the voiceover
The final audio is mastered and delivered as a playable URL

Brand context is automatically fetched from /.well-known/brand.json on your brand's domain when available.

Protocol details

AdCP version: 3.0.0-rc.2
Transport: MCP over Streamable HTTP
Session management: Stateful (session ID in mcp-session-id header)
Context tracking: Each response includes a context_id for conversation continuity
Agent discovery: /.well-known/adagents.json
Spec: docs.adcontextprotocol.org