webchat

"> webchat

NO CHAT

MODEL

MODE ?

[= USERS =]

Create accounts and control each user's role.

user	role	credits	created

[= CREDITS =]

Set global defaults and provider-by-model cost multipliers.

provider	model	mult	unknown

[= GLOBAL PROMPTS =]

Manage shared presets visible to everyone. Personal presets remain owned by their users.

[= KEYS =]

Store encrypted upstream credentials and issue OpenAI-compatible access keys.

id	provider	label	mult	masked

name	prefix	created	last used	tokens

/v1 setup — OpenAI-compatible. Base URL: https://YOUR-HOST/v1 · Auth: Authorization: Bearer <key>

{
  "$schema": "https://opencode.ai/config.json",
  "provider": {
    "webchat": {
      "npm": "@ai-sdk/openai-compatible",
      "name": "webchat",
      "options": { "baseURL": "https://YOUR-HOST/v1", "apiKey": "sk-..." },
      "models": { "qwen3.8-flash-infinite": { "name": "qwen3.8-flash-infinite" } }
    }
  }
}

# then:  opencode --model webchat/qwen3.8-flash-infinite
# OpenAI clients:  base_url="https://YOUR-HOST/v1", api_key="sk-..."
# Model ids are the same as the website picker; API keys default to NO system prompt.

[= MCP =]

Servers give the model tools such as file reads and commands through the app's 5-round tool loop. Stdio commands only; failures fall through silently.

name	command	models	on

[= FILTER =]

Block configured words before any request reaches an upstream provider.

Admin \u00b7 Custom models

Add extra model ids served through the commandcode key. The two temporary models ship by default: z-ai/glm-5.3-flash, deepseek/deepseek-v4.1-flash. Custom models are admin-only unless you untick the box.

id	model_id	name	admin_only

DEFAULT PROVIDERAuto = cheapest-first chain with failover. A picked provider = that single backend; its displayed backend multiplier applies. DEFAULT MODELSets the model used for chat and determines whether image or audio attachments are allowed. DEFAULT MODEProxy optimizes, deduplicates, and can fail over. Raw sends directly to the selected provider without those protections. SYSTEM PROMPTDefault = standard behavior, tools controlled by the TOOLS tab. Empty = no prompt and no tools at all (fastest, best for caching). A preset = prepended every turn (billed as input tokens).

[= YOUR API KEYS =]

Personal OpenAI-compatible keys — usable with opencode, curl, or any OpenAI client. Base URL: https://YOUR-HOST/v1 · Auth: Authorization: Bearer <key>. Each key can have its own system prompt, model allowlist and token limit; blank = no limit.

name	prefix	created	last used	tokens

Tip: paste this into the KEYS tab of any admin panel you run — the key works against https://YOUR-HOST/v1.