mcp server

sempera-ai

Skills, Claude Code and Codex plugin and Cursor rules that teach AI coding agents to build and operate Sempera phone agents

Sempera-AI/sempera-aicommunitylistedunreviewed

Use tool

Set up

Runs on your computer. Pick your agent and operating system.

Before you start

Node.js 20 or newer

~/Library/Application Support/Claude/claude_desktop_config.json

{
  "mcpServers": {
    "sempera-ai-sempera-ai": {
      "command": "npx",
      "args": [
        "-y",
        "@sempera/ai"
      ]
    }
  }
}

same format Cursor, GitHub Copilot CLI, Junie, JetBrains AI Assistant, Augment (Auggie CLI), Factory Droid, Kiro, Devin CLI +12

Values in angle brackets are placeholders. Restart Claude Desktop after editing its config.Claude Desktop docs

README

synced Nov 9, 2026

Sempera for AI coding agents

Documentation | MCP guide | OpenAPI document | MIT licence

Skills, a plugin and editor rules that teach an AI coding agent to build and operate Sempera phone agents: create and publish an agent, place a call or send a text, read conversations, set up a number and debug a call.

Everything here uses only the public API at https://api.sempera.org (contract: /openapi.json) with your workspace API key.

What is inside

Path	What it is
`skills/sempera-build-agent`	Create an agent, edit its draft, try it with a typed preview, publish
`skills/sempera-send-call`	Place one real call or send one real text, with consent and confirmation
`skills/sempera-read-conversations`	List and read calls and conversations with transcripts
`skills/sempera-setup-phone-number`	Check, bind and verify a number you own
`skills/sempera-debug-call`	Find out why a call, a tool or a webhook went wrong
`.claude-plugin/plugin.json`	Plugin manifest for Claude Code
`.codex-plugin/plugin.json`	Plugin manifest for Codex
`.mcp.json`	Bundles the `@sempera/mcp` server (read tools only)
`rules/sempera.mdc`	A rule file for Cursor-style editors

Install

Create an API key in the Sempera App under Settings, API keys. Give it only the scopes you want your coding agent to have.

Claude Code

# From npm
npm install @sempera/ai
claude --plugin-dir ./node_modules/@sempera/ai

# Or from a clone of this repository
claude --plugin-dir ./sempera-ai

The plugin asks for your API key once and passes it to the bundled MCP server. Or copy a skill folder from skills/ into ~/.claude/skills/ to use one skill alone.

Codex

Add this directory as a plugin (it carries .codex-plugin/plugin.json), or copy the folders from skills/ into your Codex skills directory. Set SEMPERA_API_KEY in the environment the bundled server runs in.

Cursor and other editors

Copy rules/sempera.mdc to .cursor/rules/sempera.mdc in your project. For the MCP server, add it to your client's config:

{
  "mcpServers": {
    "sempera": {
      "command": "npx",
      "args": ["-y", "@sempera/mcp"],
      "env": { "SEMPERA_API_KEY": "<your API key>" }
    }
  }
}

Quick start

Ask your coding agent, in plain words:

"List my Sempera agents and show the last five calls of the first one." (read-only)
"Build a Sempera agent for Northwind Barbers that books haircuts, and preview it." (stops before publishing)
"Why did call call-123 end early?" (reads the transcript and tool timeline)

Safety

Writes that reach people are never defaulted. A call or text needs the agent, the number and a consent sentence from you, and the agent asks you to confirm first.
Publishing, binding a number and deleting need your explicit yes.
The bundled MCP server starts read-only. Turning on send_call and send_sms (--allow-writes) is your decision, in your config.
The skills tell the agent to keep the API key and call transcripts out of files, commits and tickets.
The API still enforces calling hours, the do-not-contact list, opt-outs, caps and billing, whatever the agent asks.

Test

npm test

The tests mock the HTTP layer: they replay every curl example in the skills against a fake API that knows the public routes, and check the manifests and rules. They prove the files are consistent, not live API behaviour.