ClawCall: Let Your Agent Make Real Phone Calls

Where this fits

You're building an agent that helps users book appointments, check business hours, or resolve support tickets — but the business only answers by phone. Your agent can't call, so it stalls. ClawCall bridges that gap: when a user says "call this restaurant" or "book by phone," your agent places a confirmed outbound call, handles the conversation, and returns a transcript, summary, recording, and cost — all from a single OpenClaw skill.

Why agents benefit

• Confirmed outbound calls: The skill verifies the call connects and the conversation completes, giving your agent a reliable completion signal — not just a “call attempted” status.
• Real business number discovery: It finds public phone numbers for businesses when the user only provides a name or address, so your agent can act on partial information.
• Inbound call handling: ClawCall can act as an AI receptionist, receiving calls and routing or capturing messages — useful for agents that manage a phone line.
• Scheduled calls: Register a callback or appointment time; the skill triggers the call at the designated moment, enabling proactive follow-ups.
• Full telemetry: Transcripts, summaries, recordings, and cost data come back structured, so your agent can log, analyze, or present results to the user.

Practical scenarios

Booking a table at a busy restaurant

User asks: "Call Mario's Pizzeria and reserve a table for four at 7 PM." Your agent uses ClawCall to find the number, place the call, navigate the phone tree (with human-like speech), confirm the reservation, and return a summary with the confirmation ID.

Handling customer service calls

A user wants to dispute a charge with their bank. Your agent calls customer service, IDs the user from stored context, explains the issue, and collects the resolution — then emails the user a transcript as proof.

AI receptionist for a small business

Your agent monitors an inbound line using ClawCall. When a client calls, your agent answers, takes a message, schedules a callback, and updates the business's CRM — all without human intervention.

Add it to your agent workflow

Install the skill in your OpenClaw environment:

npx skills add ustczz/clawcall-ai-phone-calls --skill clawcall-ai-phone-calls

After installation, set the CLAW_TOKEN (or run the register command to auto-generate one). A minimal usage example inside an agent prompt:

User: "Call Acme Plumbing and ask if they can fix a leak today."
Agent: [calls ClawCall with the business name and request, receives transcript & summary, replies]

The skill accepts international E.164 numbers; for mainland China use the companion skill ai-calls-china-phone.

Compare before adopting

Before wiring this permanently, compare against alternatives:

• Quality signals: Check the repo's OpenClaw compatibility, recent commits, and any issues about call reliability or voice quality.
• Maintenance freshness: The skill is early (0 stars) — verify the author updates it for API changes.
• Alternatives: Evaluate other phone skills on OpenAgentSkill, or direct integrations like Twilio or Vonage if you need custom concurrency or carrier control.
• Workflow fit: Test whether ClawCall's CLAW_TOKEN setup and API base override align with your deployment (e.g., private server).

For a quick evaluation, see the ClawCall skill listing on OpenAgentSkill for community notes and usage patterns.

Why it is worth tracking

ClawCall is young but addresses a persistent gap: most agent frameworks stop at text and web actions. Its ability to find public numbers, make confirmed calls, handle inbound traffic, and return structured transcript data makes it a valuable building block for voice-enabled agents. Watch for adoption signals — stars, forks, and example agents built on it — and re-evaluate once a few real-world deployments appear, because early feedback will shape API stability and reliability.