Famulor Skill - Famulor

Famulor Skill teaches your AI client how to design assistants and work with Famulor. It includes prompt templates, channel guidance, testing scenarios, and references for the public API, MCP, CLI, and TypeScript SDK. View the repository · Download the latest release · Detailed installation guide

The skill provides instructions. An MCP connection provides authorized workspace access. Installing the skill alone does not connect your account or expand an existing connection’s permissions. You can draft prompts without a connection; reading or changing workspace data requires one.

Choose your starting point

You want to…	Start with
Build a voice receptionist or booking assistant	Assistant design and voice templates
Adapt an assistant to WhatsApp, web chat, or email	Channel guidance and prompt examples
Improve conversation quality	Testing and evaluation
Find an endpoint, CLI command, or SDK method	API and SDK guide
Prepare campaigns, knowledge, or automations	Workflow examples and MCP use cases

Learn from complete examples

Use the worked-example library to see a complete assistant design inspired by the public Famulor demos.

Example	What it teaches
Hotel front desk	Greeting, hotel knowledge, actual availability, booking confirmation and recovery
Drive-thru	Correcting a basket, checking its quote and submitting once
Order support	Shared business policy with voice, WhatsApp, webchat and email adaptations
Lead qualification	Useful discovery, appropriate next steps and respectful opt-out handling

Each example includes a filled brief, complete prompt, setting placement, required runtime tools, an illustrative conversation and acceptance checks. The assistant deliverable shows how an agent should present the finished proposal and verification status. The businesses and example outcomes are fictional: gallery role-play does not prove that a booking or order happened. Connect real tools and verify their results before making those claims to customers.

Install

For coding agents that support Agent Skills, run:

npx skills add bekservice/Famulor-Skill --skill famulor-skill

Choose your client in the installer. It installs into the current project by default; add -g for a global installation. Review the package before installing it. For workspace actions, add a remote Streamable HTTP MCP connection and complete OAuth:

https://app.famulor.io/mcp

Use the domain shown in Settings → API & MCP for a white-label workspace. Your plan must include Connect AI / MCP, and your workspace role and approved scopes must allow the requested action. A focused URL such as https://app.famulor.io/mcp?toolsets=assistants,calls limits tool discovery without granting extra permissions.

Install the bundled plugin from the repository’s marketplace:

/plugin marketplace add bekservice/Famulor-Skill
/plugin install famulor@famulor

Open /mcp and authenticate Famulor. This package includes both the skill and MCP configuration. If you installed only the skill with npx skills, add the connection separately:

claude mcp add --transport http famulor https://app.famulor.io/mcp

Download famulor-skill.zip from the latest release.
Open Customize → Skills → + → Create skill → Upload a skill and upload that ZIP. Do not upload the whole repository ZIP.
Enable the skill. Your account or organization must permit skills and code execution.
Connect Famulor under Customize → Connectors. Use a custom connector with the MCP URL for the tools you need.
Start a new conversation and ask Claude to use Famulor Skill.

The Famulor Assistants & History directory connector provides read-only access. It cannot save prompts or send messages. See Claude’s skill upload guide if your controls differ.

After installing the skill, configure and verify MCP:

codex mcp add famulor --url https://app.famulor.io/mcp
codex mcp login famulor
codex mcp get famulor

Restart the session and ask Codex to use $famulor-skill. IDE clients can also add the connection under MCP servers → Add server → Streamable HTTP.

Choose Cursor in the skill installer. Add the remote URL in Cursor’s MCP settings, or use this project configuration in .cursor/mcp.json:

{
  "mcpServers": {
    "famulor": {
      "url": "https://app.famulor.io/mcp"
    }
  }
}

Authenticate when prompted and verify that the Famulor tools are enabled.

Install the extension, which bundles the skill and MCP configuration:

gemini extensions install https://github.com/bekservice/Famulor-Skill

Review the requested capabilities. Run /mcp auth famulor if authentication is requested, then /extensions list to check the installation.

For other compatible clients, use the installer or copy the complete skills/famulor-skill/ folder, including its references, scripts, and templates. Copying only SKILL.md omits the supporting guides. The release also provides famulor.skill for compatible importers.

Verify with a read-only request

Use Famulor to list my assistants. Tell me which workspace is connected. Do not change anything.

Then try a focused review:

Use Famulor Skill to review my support assistant for website text chat. Propose a revised prompt and three test conversations. Do not save or send anything.

A downloaded package or completed OAuth screen does not prove that a tool call succeeded. Confirm that the client returned actual data from the intended workspace. On a read-only connection, review and draft workflows work; saving changes requires separate write access.

Prompting that matches each channel

Keep business policy, trusted sources, and action rules in the shared system prompt. Use Additional channel instructions for tone, length, and layout when Manual channel responses is included in your plan.

Voice: concise spoken turns, one clear question, an appropriate greeting, and real transfer or booking tools.
WhatsApp: compact messages, human takeover, the current service window, and approved templates for authorized outreach when required.
Web chat: readable links, pre-chat context, useful written answers, and a test of the actual widget.
Email: complete body text, the right thread and signature behavior, and a clear distinction between a draft and a sent reply.

A prompt cannot connect a calendar, create a tool, grant permissions, or prove a delivery. The skill guides the client to inspect configuration, resolve real resource IDs, use supported actions, and verify the result. Start with the use-case requests, then test the channel your customers will actually use.

Update and troubleshoot

For an installer-managed skill, run npx skills check, then npx skills update. For a manually uploaded Claude skill, upload the latest release ZIP. Keep the MCP connection separately authorized; removing local files does not necessarily revoke OAuth access.

Symptom	Next step
Skill installed, no tools	Add the MCP connection, authenticate, and restart the client session.
Only 11 read-only tools	You connected the directory profile. Use a separately authorized custom connection for broader actions.
Cannot save an assistant	Check the workspace, role, approved scopes, and plan.
A write timed out	Read current state before retrying; the first action may have completed.
Installation controls differ	Follow the linked maintained installation guide and your client’s official documentation.

Use the repository’s issues for reproducible skill or documentation problems. Remove customer data and credentials before sharing an example. For account-specific help, use Famulor support.

​Choose your starting point

​Learn from complete examples

​Install

​Verify with a read-only request

​Prompting that matches each channel

​Update and troubleshoot

Choose your starting point

Learn from complete examples

Install

Verify with a read-only request

Prompting that matches each channel

Update and troubleshoot