Skip to main content

Documentation Index

Fetch the complete documentation index at: https://docs.getpatter.com/llms.txt

Use this file to discover all available pages before exploring further.

OpenClaw is a self-hosted AI assistant gateway that lets one or more LLM-backed agents act on your behalf across WhatsApp, Telegram, Discord, iMessage, and more. Once you install the Patter skill, your OpenClaw agent can place outbound calls, receive inbound calls, and use the full Patter SDK without any further setup on the OpenClaw side.

​
Install

Patter skills work in every agent harness that consumes the Anthropic Agent Skills standard — OpenClaw included. Install with one command: 
# Install all five Patter skills
npx skills add patterai/skills

# Or install just the one you need
npx skills add patterai/skills --skill build-voice-agent
The CLI auto-detects OpenClaw on your machine and creates symlinks from ~/.openclaw/skills/ (or your CLAWD_HOME override) into the canonical ~/.agents/skills/ directory. Restart your Gateway and the agent will pick the skills up at session start.
Pin to a specific Patter SDK version for reproducibility:
npx skills add patterai/skills#v0.6.2 --skill build-voice-agent

​
Available skills

SkillWhat it teaches the agent
setup-patterWalk the user through installing Patter and obtaining provider/carrier API keys, with validation.
build-voice-agentBuild a voice agent — pick between OpenAI Realtime / ElevenLabs ConvAI / STT→LLM→TTS Pipeline.
configure-telephonyWire up Twilio, Telnyx, or Plivo — phone numbers, webhooks, tunnels, AMD, voicemail drop.
add-tools-and-handoffsCustom tools, transfer_call, end_call, output guardrails.
inspect-calls-and-metricsLive dashboard, CallMetrics, cost tracking, CSV/JSON export.
Every skill ships full Python and TypeScript code examples side-by-side. Your OpenClaw agent can pick the runtime that matches the rest of your project.

​
Use Patter as an MCP server (alternative path)

If you want OpenClaw to call Patter through the Model Context Protocol instead of (or in addition to) the skills, register the Patter MCP server with one command: 
openclaw mcp set patter '{
  "command": "npx",
  "args": ["-y", "getpatter-mcp"],
  "env": {
    "TWILIO_ACCOUNT_SID": "AC...",
    "TWILIO_AUTH_TOKEN": "...",
    "OPENAI_API_KEY": "sk-..."
  }
}'
The Patter MCP server is hosted at PatterAI/patter-mcp. Once registered, tools surface to your OpenClaw agent under names like patter__make_call, patter__get_transcript, and patter__list_calls.
Allowlist outbound calls. OpenClaw’s before_tool_call.requireApproval hook applies to native plugins but not to MCP tools. If you don’t want your agent dialing arbitrary numbers without confirmation, configure tools.elevated: ["patter__make_call"] in your OpenClaw config.yaml, or wrap the tool in a small native plugin that gates it behind user approval.

​
Quick start: outbound call from OpenClaw

Once skills are installed and your provider/carrier keys are in ~/.openclaw/.env, ask your OpenClaw agent something like:
“Call my plumber at +15550001234 and ask if they have time tomorrow morning. Then text me the summary.”
The agent loads build-voice-agent automatically (its description triggers on phrases like “call”, “phone”, “voicemail”), writes a small Python or TypeScript script using the Patter SDK, runs it, and reports the transcript back to you in the OpenClaw thread. If you’d prefer to skip the per-call SDK script and have the agent use the MCP server tools directly, ensure patter is in your tools.allow list: 
# ~/.openclaw/config.yaml
tools:
  allow:
    - bundle-mcp
    - patter__*

​
What’s next

  • Read the Concepts page to understand Patter’s three modes (Realtime / ConvAI / Pipeline) and which one matches your latency / cost trade-off.
  • Get a phone number: Twilio console or Telnyx portal.
  • Provision your provider keys: the setup-patter skill walks the agent (and you) through every provider dashboard, one at a time, with curl validation for each key.
  • Browse the skill content: https://github.com/PatterAI/skills.

​
Need help?