One API for every model.
Route requests across providers, track every token and dollar, and switch models without changing your code.
Pay per token. No subscription.
Models from leading AI labs
OpenAI
Claude
Gemini
Grok
Muse
DeepSeek
Kimi
Qwen
GLM
MiniMax
Dashboard
Everything in one dashboard.
Usage, spend, keys and requests for every model, in real time.
Unified API
One request. Any model.
Switch models by changing one string. Same request shape, same SDK, same billing balance.
- Same OpenAI-compatible request body across every lab
- Live per-token pricing and context windows
- Pin a specific model or pass auto for smart routing
import OpenAI from "openai";
const client = new OpenAI({
baseURL: "https://api.vichar.io/v1",
apiKey: process.env.VICHAR_API_KEY,
});
const resp = await client.chat.completions.create({
model: "vichar-opus-5.5",selected messages: [{ role: "user", content: "Hello" }],
});
console.log(resp.choices[0].message.content);Context
1M
Input / 1M
$4.00
Output / 1M
$20.00
Controls
Built for production.
Spend limits, expiring keys, smart routing and caching come standard.
Spend limits per key
Cap any API key at a fixed total or a recurring daily, weekly, or monthly budget, so a runaway job cannot drain your balance.
Key expiry
Give keys a lifetime for contractors, demos, or rotating services. They stop working at the time you set.
Smart routing
Send requests to auto and Vichar picks one of your configured models based on how hard the prompt is.
Prompt caching
Repeated prompt prefixes hit provider caches, and cached input tokens are billed at the model's lower cached rate.
Shared workspace
Every key in the workspace draws from one prepaid balance, with roles and per-member budgets.
Rate limits that grow
New accounts start at 600 requests per minute. Limits rise automatically as your account ages and spends.
Reliability
Retries and failover, built in.
When a provider errors or times out, the request is retried and rerouted to another available route for the same model. Failed attempts are not billed.
- Automatic retry on 5xx and upstream timeouts
- Failed attempts never deduct from your balance
- Send x-no-fallback: true when you want strict single-route calls
Drop-in migration
Switch in two lines.
Already using the OpenAI SDK? Point baseURL at Vichar, swap the API key, and keep the rest of your application untouched.
- Works with streaming, tool calling, and structured outputs
- Also supports the Anthropic /v1/messages format
import OpenAI from "openai";const client = new OpenAI({- apiKey: process.env.OPENAI_API_KEY,+ baseURL: "https://api.vichar.io/v1",+ apiKey: process.env.VICHAR_API_KEY,});const response = await client.chat.completions.create({ model: "vichar-sonnet-5.5", messages: [{ role: "user", content: "Hello" }],});Integrations
Works with your stack.
Any SDK or coding tool that accepts an OpenAI-compatible base URL works with Vichar.
Catalogue
Frontier models, ready to call.
Transparent per-token pricing across reasoning, coding, and multimodal models.
| Model | Lab | Context | Input / 1M | Output / 1M |
|---|---|---|---|---|
| Claude Opus 5.5 vichar-opus-5.5 | Anthropic | 1M | $4.00 | $20.00 |
| GPT-5.6 vichar-gpt-5.6 | OpenAI | 1.1M | $2.00 | $12.00 |
| Gemini 3.1 Pro vichar-gemini-3.1-pro | Google | 1M | $2.00 | $12.00 |
| Grok 4.7 vichar-grok-4.7 | xAI | 500k | $2.00 | $6.00 |
| Claude Sonnet 5.5 vichar-sonnet-5.5 | Anthropic | 1M | $2.00 | $10.00 |
| DeepSeek V4 Pro vichar-deepseek-v4-pro | DeepSeek | 1M | $0.66 | $1.98 |
Privacy
Privacy and control by default.
By default, prompts and responses are never written to our database. We keep the metadata needed to bill you, nothing more.
- Keys hashed at rest
- SHA-256
- API keys are stored as one-way hashes. The full key is shown once, at creation.
- Limits and instant revocation
- per key
- Cap any key's spend or give it an expiry, and revoke it in one click if it leaks.
- Encrypted in transit
- TLS 1.2+
- The API is HTTPS only, between your app, Vichar, and the model providers.
Details in the Privacy Policy.
Pricing
Model rates plus 5% on top-ups. That's the whole bill.
- No subscription, seat fees, or monthly minimum
- Spend limits per API key and optional auto top-up
- Downloadable invoices for every top-up
Inference usage
Listed rate
0% per-token markup
Credit top-up
5% fee
$20 top-up = $21 total
Failed requests and upstream errors are never billed.
Frequently asked questions
Is Vichar compatible with the OpenAI SDK?+
Yes. Point your client at https://api.vichar.io/v1 and use a Vichar API key. Your existing code keeps working.
How is usage billed?+
Per token at the model's listed rate, deducted from prepaid credits. Top-ups carry a 5% platform fee and a $20 minimum.
Do you store my prompts?+
No. By default we keep request metadata only: tokens, cost, latency, and which provider served the request.
What happens if a provider goes down?+
Requests automatically reroute to another provider serving the same model when one is available.
Are there rate limits?+
New accounts get 600 requests per minute and 100 concurrent requests. Limits rise as accounts grow.
Can I set spending limits?+
Yes. Every API key can carry its own spend limit, so a runaway workload cannot drain your balance.