elevenlabs-agents

Name: elevenlabs-agents
Author: jezweb

1.5kinstalls96forks

How to install elevenlabs-agents?

jezweb/claude-skills/elevenlabs-agents

$npx -y skills add jezweb/claude-skills --skill elevenlabs-agents

Installs into the current project.

Summary

Build conversational AI voice agents on the ElevenLabs platform. Configure agent + tools + knowledge base, integrate SDK (React / React Native / Swift / JS / server-side), test, deploy. Use whenever the user mentions ElevenLabs, building a voice agent, an AI phone system, an AI r

Files · 1

View on GitHub

SKILL.md

1	# ElevenLabs Agent Builder
2
3	Build a production-ready conversational AI voice agent. Produces a configured agent with tools, knowledge base, and SDK integration.
4
5	## Packages
6
7	```bash
8	npm install @elevenlabs/react # React SDK
9	npm install @elevenlabs/client # JavaScript SDK (browser + server)
10	npm install @elevenlabs/react-native # React Native SDK
11	npm install @elevenlabs/elevenlabs-js # Full API (server only)
12	npm install -g @elevenlabs/agents-cli # CLI ("Agents as Code")
13	```
14
15	DEPRECATED: `@11labs/react`, `@11labs/client` -- uninstall if present.
16
17	Server-only warning: `@elevenlabs/elevenlabs-js` uses Node.js `child_process` and won't work in browsers. Use `@elevenlabs/client` for browser environments, or create a proxy server.
18
19	## Workflow
20
21	### Step 1: Create Agent via Dashboard or CLI
22
23	Dashboard: https://elevenlabs.io/app/conversational-ai -> Create Agent
24
25	CLI (Agents as Code):
26	```bash
27	elevenlabs agents init
28	elevenlabs agents add "Support Bot" --template customer-service
29	# Edit agent_configs/support-bot.json
30	elevenlabs agents push --env dev
31	```
32
33	Templates: `default`, `minimal`, `voice-only`, `text-only`, `customer-service`, `assistant`.
34
35	Configure:
36	- Voice -- Choose from 5000+ voices or clone
37	- LLM -- GPT, Claude, Gemini, or custom
38	- System prompt -- Use the 6-component framework below
39	- First message -- What the agent says when conversation starts
40
41	### Step 2: Write the System Prompt
42
43	Use the 6-component framework for effective agent prompts:
44
45	1. Personality -- who the agent is:
46	```
47	You are [NAME], a [ROLE] at [COMPANY].
48	You have [EXPERIENCE]. Your traits: [LIST TRAITS].
49	```
50
51	2. Environment -- communication context:
52	```
53	You're communicating via [phone/chat/video].
54	Consider [environmental factors]. Adapt to [context].
55	```
56
57	3. Tone -- speech patterns and formality:
58	```
59	Tone: Professional yet warm. Use contractions for natural speech.
60	Avoid jargon. Keep responses to 2-3 sentences. Ask one question at a time.
61	```
62
63	4. Goal -- objectives and success criteria:
64	```
65	Primary Goal: Resolve customer issues on the first call.
66	Success: Customer verbally confirms issue is resolved.
67	```
68
69	5. Guardrails -- boundaries and ethics:
70	```
71	Never: provide medical/legal/financial advice, share confidential info.
72	Always: verify identity before account access, document interactions.
73	Escalation: customer requests manager, issue beyond knowledge base.
74	```
75
76	6. Tools -- available functions and when to use them:
77	```
78	1. lookup_order(order_id) -- Use when customer mentions an order.
79	2. transfer_to_supervisor() -- Use when issue requires manager approval.
80	Always explain what you're doing before calling a tool.
81	```
82
83	### Step 3: Add Tools
84
85	Client-side tools (run in browser):
86
87	```typescript
88	const clientTools = {
89	updateCart: {
90	description: "Add or remove items from the shopping cart",
91	parameters: z.object({
92	action: z.enum(['add', 'remove']),
93	item: z.string(),
94	quantity: z.number().min(1)
95	}),
96	handler: async ({ action, item, quantity }) => {
97	const cart = getCart();
98	action === 'add' ? cart.add(item, quantity) : cart.remove(item, quantity);
99	return { success: true, total: cart.total, items: cart.items.length };
100	}
101	},
102	navigate: {
103	description: "Navigate user to a different page",
104	parameters: z.object({ url: z.string().url() }),
105	handler: async ({ url }) => { window.location.href = url; return { success: true }; }
106	}
107	};
108	```
109
110	Server-side tools (webhooks):
111
112	```json
113	{
114	"name": "get_weather",
115	"description": "Fetch current weather for a city",
116	"url": "https://api.weather.com/v1/current",
117	"method": "GET",
118	"parameters": {
119	"type": "object",
120	"properties": {
121	"city": { "type": "string", "description": "City name" }
122	},
123	"required": ["city"]
124	},
125	"headers": {
126	"Authorization": "Bearer {{secret__weather_api_key}}"
127	}
128	}
129	```
130
131	Use `{{secret__key_name}}` for API keys in webhook headers -- never hardcode.
132
133	MCP Tools -- CRITICAL COMPATIBILITY NOTE:
134
135	ElevenLabs labels their MCP integration as "Streamable HTTP" but does NOT support the actual MCP 2025-03-26 Streamable HTTP spec (SSE responses). ElevenLabs expects:
136
137	- Plain JSON responses (`application/json`), NOT SSE (`text/event-stream`)
138	- Protocol version `2024-11-05`, NOT `2025-03-26`
139	- Simple JSON-RPC over HTTP with direct JSON responses
140
141	What does NOT work:
142	- Official MCP SDK's `createMcpHandler` (returns SSE)
143	- Cloud

People who liked this also viewed

6 picks

Type

elevenlabs-agents

1.5kinstalls96forks

View on GitHub

How to install elevenlabs-agents?

jezweb/claude-skills/elevenlabs-agents

$npx -y skills add jezweb/claude-skills --skill elevenlabs-agents

Installs into the current project.

Summary

Files · 1

View on GitHub

SKILL.md

1	# ElevenLabs Agent Builder
2
3	Build a production-ready conversational AI voice agent. Produces a configured agent with tools, knowledge base, and SDK integration.
4
5	## Packages
6
7	```bash
8	npm install @elevenlabs/react # React SDK
9	npm install @elevenlabs/client # JavaScript SDK (browser + server)
10	npm install @elevenlabs/react-native # React Native SDK
11	npm install @elevenlabs/elevenlabs-js # Full API (server only)
12	npm install -g @elevenlabs/agents-cli # CLI ("Agents as Code")
13	```
14
15	DEPRECATED: `@11labs/react`, `@11labs/client` -- uninstall if present.
16
17	Server-only warning: `@elevenlabs/elevenlabs-js` uses Node.js `child_process` and won't work in browsers. Use `@elevenlabs/client` for browser environments, or create a proxy server.
18
19	## Workflow
20
21	### Step 1: Create Agent via Dashboard or CLI
22
23	Dashboard: https://elevenlabs.io/app/conversational-ai -> Create Agent
24
25	CLI (Agents as Code):
26	```bash
27	elevenlabs agents init
28	elevenlabs agents add "Support Bot" --template customer-service
29	# Edit agent_configs/support-bot.json
30	elevenlabs agents push --env dev
31	```
32
33	Templates: `default`, `minimal`, `voice-only`, `text-only`, `customer-service`, `assistant`.
34
35	Configure:
36	- Voice -- Choose from 5000+ voices or clone
37	- LLM -- GPT, Claude, Gemini, or custom
38	- System prompt -- Use the 6-component framework below
39	- First message -- What the agent says when conversation starts
40
41	### Step 2: Write the System Prompt
42
43	Use the 6-component framework for effective agent prompts:
44
45	1. Personality -- who the agent is:
46	```
47	You are [NAME], a [ROLE] at [COMPANY].
48	You have [EXPERIENCE]. Your traits: [LIST TRAITS].
49	```
50
51	2. Environment -- communication context:
52	```
53	You're communicating via [phone/chat/video].
54	Consider [environmental factors]. Adapt to [context].
55	```
56
57	3. Tone -- speech patterns and formality:
58	```
59	Tone: Professional yet warm. Use contractions for natural speech.
60	Avoid jargon. Keep responses to 2-3 sentences. Ask one question at a time.
61	```
62
63	4. Goal -- objectives and success criteria:
64	```
65	Primary Goal: Resolve customer issues on the first call.
66	Success: Customer verbally confirms issue is resolved.
67	```
68
69	5. Guardrails -- boundaries and ethics:
70	```
71	Never: provide medical/legal/financial advice, share confidential info.
72	Always: verify identity before account access, document interactions.
73	Escalation: customer requests manager, issue beyond knowledge base.
74	```
75
76	6. Tools -- available functions and when to use them:
77	```
78	1. lookup_order(order_id) -- Use when customer mentions an order.
79	2. transfer_to_supervisor() -- Use when issue requires manager approval.
80	Always explain what you're doing before calling a tool.
81	```
82
83	### Step 3: Add Tools
84
85	Client-side tools (run in browser):
86
87	```typescript
88	const clientTools = {
89	updateCart: {
90	description: "Add or remove items from the shopping cart",
91	parameters: z.object({
92	action: z.enum(['add', 'remove']),
93	item: z.string(),
94	quantity: z.number().min(1)
95	}),
96	handler: async ({ action, item, quantity }) => {
97	const cart = getCart();
98	action === 'add' ? cart.add(item, quantity) : cart.remove(item, quantity);
99	return { success: true, total: cart.total, items: cart.items.length };
100	}
101	},
102	navigate: {
103	description: "Navigate user to a different page",
104	parameters: z.object({ url: z.string().url() }),
105	handler: async ({ url }) => { window.location.href = url; return { success: true }; }
106	}
107	};
108	```
109
110	Server-side tools (webhooks):
111
112	```json
113	{
114	"name": "get_weather",
115	"description": "Fetch current weather for a city",
116	"url": "https://api.weather.com/v1/current",
117	"method": "GET",
118	"parameters": {
119	"type": "object",
120	"properties": {
121	"city": { "type": "string", "description": "City name" }
122	},
123	"required": ["city"]
124	},
125	"headers": {
126	"Authorization": "Bearer {{secret__weather_api_key}}"
127	}
128	}
129	```
130
131	Use `{{secret__key_name}}` for API keys in webhook headers -- never hardcode.
132
133	MCP Tools -- CRITICAL COMPATIBILITY NOTE:
134
135	ElevenLabs labels their MCP integration as "Streamable HTTP" but does NOT support the actual MCP 2025-03-26 Streamable HTTP spec (SSE responses). ElevenLabs expects:
136
137	- Plain JSON responses (`application/json`), NOT SSE (`text/event-stream`)
138	- Protocol version `2024-11-05`, NOT `2025-03-26`
139	- Simple JSON-RPC over HTTP with direct JSON responses
140
141	What does NOT work:
142	- Official MCP SDK's `createMcpHandler` (returns SSE)
143	- Cloud

People who liked this also viewed

6 picks

Type