self-setup

name	self-setup
description	Configure LLM providers and models through conversation — add API keys, switch models, list available options.
metadata	{"adclaw":{"emoji":"⚙️"}}

Self-Setup — Provider & Model Configuration

This skill lets you configure LLM providers and models through conversation. You have access to your own management API at http://localhost:8088/api/.

---

When to Use

• User asks to add, change, or configure an LLM provider
• User asks to switch models
• User provides an API key for a provider
• User asks "what models are available?"
• User says something like "use GPT-5" or "switch to Claude"

---

API Reference

All commands use curl via your shell tool. Base URL: http://localhost:8088/api

List All Providers

curl -s http://localhost:8088/api/models | python3 -m json.tool

Get Active Model

curl -s http://localhost:8088/api/models/active | python3 -m json.tool

Set Active Model

curl -s -X PUT http://localhost:8088/api/models/active \
  -H "Content-Type: application/json" \
  -d '{"provider_id": "PROVIDER_ID", "model": "MODEL_ID"}'

Configure Provider API Key

curl -s -X PUT http://localhost:8088/api/models/PROVIDER_ID/config \
  -H "Content-Type: application/json" \
  -d '{"api_key": "USER_API_KEY"}'

Add Custom Provider

For providers not in the built-in list (Together AI, Groq, Fireworks, DeepInfra, etc.):

curl -s -X POST http://localhost:8088/api/models/custom-providers \
  -H "Content-Type: application/json" \
  -d '{
    "id": "provider-id",
    "name": "Provider Name",
    "default_base_url": "https://api.example.com/v1",
    "api_key_prefix": "sk-",
    "models": [
      {"id": "model-id", "name": "Model Display Name"}
    ]
  }'

Test Provider Connection

curl -s -X POST http://localhost:8088/api/models/PROVIDER_ID/test \
  -H "Content-Type: application/json" \
  -d '{"api_key": "OPTIONAL_KEY", "base_url": "OPTIONAL_URL"}'

Delete Custom Provider

curl -s -X DELETE http://localhost:8088/api/models/PROVIDER_ID

Add Model to Existing Provider

curl -s -X POST http://localhost:8088/api/models/PROVIDER_ID/models \
  -H "Content-Type: application/json" \
  -d '{"id": "model-id", "name": "Model Name"}'

---

Built-in Providers

ID	Name	Base URL	Key prefix
openrouter	OpenRouter	openrouter.ai/api/v1	sk-or-
openai	OpenAI	api.openai.com/v1	sk-
anthropic	Anthropic	api.anthropic.com/v1	sk-ant-
aliyun-intl	Alibaba Coding (International)	coding-intl.dashscope.aliyuncs.com/v1	sk-sp
aliyun-codingplan	Alibaba Coding (China)	coding.dashscope.aliyuncs.com/v1	sk-sp
ollama	Ollama (local)	localhost:11434/v1	(none)

---

Common Custom Providers

Name	Base URL	Key prefix
Together AI	https://api.together.xyz/v1	(none)
Groq	https://api.groq.com/openai/v1	gsk_
Fireworks AI	https://api.fireworks.ai/inference/v1	fw_
DeepInfra	https://api.deepinfra.com/v1/openai	(none)
Mistral AI	https://api.mistral.ai/v1	(none)

---

Behavior Guidelines

1. Always list current providers first before making changes
2. Confirm with the user before switching models or adding providers
3. Never expose full API keys in responses — show only prefix
4. Test connection after configuring a new provider
5. If user says "use Claude" → suggest OpenRouter (one key for all) or Anthropic (direct)
6. If user says "use a free model" → suggest Ollama (local) or Aliyun-Intl (free trial)

---

Citedy MCP Configuration

To check or update the Citedy MCP tools integration:

# Check status
curl -s http://localhost:8088/api/citedy/status | python3 -m json.tool

# Save Citedy API key
curl -s -X POST http://localhost:8088/api/citedy/save-api-key \
  -H "Content-Type: application/json" \
  -d '{"api_key": "citedy_agent_..."}'

---

Channel Configuration

# List MCP clients
curl -s http://localhost:8088/api/mcp | python3 -m json.tool

# Add/update MCP client
curl -s -X POST http://localhost:8088/api/mcp \
  -H "Content-Type: application/json" \
  -d '{
    "key": "client_key",
    "name": "Client Name",
    "enabled": true,
    "transport": "streamable_http",
    "url": "https://server.example.com/mcp",
    "headers": {"Authorization": "Bearer TOKEN"}
  }'