Venarm for developers

One API.
Every route.

Build against a familiar interface while Venarm handles providers, credentials, fallbacks, limits, and observability behind the scenes.

request.js
const response = await fetch("/v1/chat/completions", {
  method: "POST",
  headers: {
    "Authorization": `Bearer ${VINEM_KEY}`,
    "Content-Type": "application/json"
  },
  body: JSON.stringify({
    model: "venarm-fast",
    reasoning_effort: "medium",
    messages,
    stream: true
  })
});

// Provider routing and fallback are automatic.
Developer API · OpenAI-compatible chat completions are available to accounts with API access enabled. Your key is shown only once when created.

Your API keys

Create a key for an IDE or app. Store it as a secret in that tool; Venarm stores only a one-way hash.

Checking account…

A revoked key stops working immediately. You can keep up to 10 active keys.

Platform

Infrastructure without the busywork

Venarm is being designed as a dependable model gateway with one policy layer and clear operational visibility.

Intelligent routing

Choose a logical model while Venarm selects an eligible provider and follows the configured fallback chain.

Credential isolation

Provider secrets remain encrypted on the server and never enter browser code, logs, or client responses.

Built-in observability

Trace attempts, latency, failures, fallback behavior, and token reporting from the Admin command center.

Compatibility

A surface your tools already understand

OpenAI-compatible chat completions and model discovery. Tool schemas supplied by your IDE are translated into function calls; your IDE runs those tools against its own project and permissions.

POST/v1/chat/completionsChat, function calls, and OpenAI-style streamed responses
GET/v1/modelsModels available to the authenticated account and plan
reasoning effort
{
  "model": "venarm-fast",
  "reasoning_effort": "high",
  "messages": [{ "role": "user", "content": "Review this change and test it." }]
}

Use reasoning_effort (or thinking_mode) with low, medium, high, or ultra. Medium is the default. Supported providers receive their native effort setting; other compatible providers receive mode-specific task guidance and a larger completion budget. Fallback attempts keep the same conversation context and effort. This changes reasoning effort, not model capability or context-window size.

OpenCode

Connect your coding workspace

Use your deployment URL and API key. OpenCode supplies its own read, write, edit, search, and terminal tools; Venarm does not access your computer's files.

opencode.json
{
  "$schema": "https://opencode.ai/config.json",
  "provider": {
    "venarm": {
      "npm": "@ai-sdk/openai-compatible",
      "name": "Venarm",
      "options": {
        "baseURL": "https://venarm.tail84d48a.ts.net/v1",
        "apiKey": "{env:VENARM_API_KEY}"
      },
      "models": {
        "venarm-fast": { "name": "Venarm Fast" },
        "venarm-pro": { "name": "Venarm Pro" }
      }
    }
  },
  "model": "venarm/venarm-fast"
}

Set VENARM_API_KEY in your environment or use OpenCode's credential command. Venarm applies the published Agent/API prompt; IDE system messages cannot replace it.

coding-guidance.md · optional
When working in this repository, first inspect its structure and existing conventions.
Make the smallest complete change that solves the request. Read relevant files before editing,
preserve unrelated user changes, and never claim a command or test succeeded unless it ran.
Use the available read, search, edit, write, and terminal tools only when they materially help.
After editing, explain the key changes and list any verification that actually completed.
Keep the repository's existing instructions and the Venarm system prompt in effect.