NO CHAT
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.
No prompts yet. Add one here in Admin → Prompts.
[= 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.
New API key: copy now, shown once
Admin \u00b7 Custom models
Add extra model ids served through the commandcode key. The three temporary models ship by default: z-ai/glm-5.3-flash, deepseek/deepseek-v4.1-flash, meta/muse-spark-1.3-contributor. Custom models are admin-only unless you untick the box.
| id | model_id | name | admin_only |
|---|