CallSphere MCP Server

A public, read-only Model Context Protocol server giving any AI assistant — Claude, Claude Desktop, Cursor, Cline, Continue, Windsurf, Zed, or ChatGPT via JSON-RPC — direct access to CallSphere's voice-AI knowledge base, industry ROI calculator, case studies for 6 verticals, and product facts.

- Endpoint: https://callsphere.ai/api/mcp
- Transport: Streamable HTTP
- Manifest: https://callsphere.ai/.well-known/mcp.json
- Docs: https://callsphere.ai/mcp
- Auth: None — free, public, read-only
- Rate limit: 120 req/min/IP

Tools

| Tool | Description |
|------|-------------|
| search_voice_ai_docs | Full-text search across 500+ articles on voice AI, MCP, Claude, agentic AI, Twilio/WebRTC, and how-to guides. |
| read_doc | Fetch the full markdown body of a single CallSphere blog post by slug. |
| calculate_voice_ai_roi | Industry-aware ROI math: monthly savings, payback period, recommended plan tier. Returns assumptions so the AI can show its work. |
| get_industry_case_study | Narrative case study for one of 6 verticals (dental, healthcare, HVAC, real estate, salon, law). |
| get_callsphere_info | Canonical lookup: pricing tiers, integrations, compliance posture, supported languages, founder contact. |

Install

Claude Code

claude mcp add --transport http callsphere https://callsphere.ai/api/mcp

Claude Desktop

Edit claude_desktop_config.json (Settings → Developer → Edit Config) and merge:

{
  "mcpServers": {
    "callsphere": {
      "command": "npx",
      "args": ["-y", "mcp-remote", "https://callsphere.ai/api/mcp"]
    }
  }
}

The mcp-remote bridge is used for compatibility with Claude Desktop builds that don't yet speak Streamable HTTP directly. Newer builds can use the simpler:

{
  "mcpServers": {
    "callsphere": {
      "url": "https://callsphere.ai/api/mcp"
    }
  }
}

Cursor

Settings → MCP → Add new MCP server, then paste:

{
  "mcpServers": {
    "callsphere": {
      "url": "https://callsphere.ai/api/mcp",
      "transport": "http"
    }
  }
}

Cline / Continue / Windsurf

Add to your MCP servers config:

"callsphere": {
  "url": "https://callsphere.ai/api/mcp",
  "type": "http"
}

ChatGPT / any HTTP client (no MCP support)

The same endpoint accepts plain JSON-RPC 2.0 POST requests:

curl -sS https://callsphere.ai/api/mcp \
  -H 'Content-Type: application/json' \
  -d '{"jsonrpc":"2.0","id":1,"method":"tools/call",
       "params":{"name":"calculate_voice_ai_roi",
                 "arguments":{"industry":"dental","monthly_calls":1200}}}'

Try these prompts

- "Search CallSphere for articles about building MCP servers with Twilio."
- "Calculate the voice-AI ROI for a 4-provider dental practice doing 1,500 calls a month — show your assumptions."
- "Show me a case study for HVAC after-hours dispatch."
- "Compare CallSphere pricing tiers and tell me which fits 8,000 calls a month."
- "What languages does CallSphere support, and is it HIPAA-ready?"

Discovery

This server publishes a machine-readable manifest at /.well-known/mcp.json so MCP registries (mcp.so, Glama.ai, etc.) can auto-discover and index it. See manifest.json in this repo for an offline copy.

What's NOT in this server (yet)

Phase 2 write tools — outbound demo calls, newsletter signup, demo booking — are intentionally paused. They would incur per-call Twilio/SES costs on what is currently free public infrastructure. They'll arrive when traffic justifies the cost, with explicit per-tool user consent and verification.

License

MIT for the install snippets and examples in this repo. The MCP server itself is hosted by CallSphere LLC; tool outputs are produced from CallSphere's own data and product configuration.

Maintained by

CallSphere LLC — Sagar Shankaran (sagar@callsphere.ai)