🧮
Bills nobody can explain
Provider invoices show totals, not reasons. InstaRoute shows the cost of every request next to what a premium model would have charged.
AI gateway & LLM router · built in India
Every AI request on the right model — with a receipt to prove it.
InstaRoute is the cost-control layer between your apps and 29 AI providers. It judges how hard each request is, sends it to the cheapest model that will do the job well, detours around failing providers and itemises what you saved — through one OpenAI- and Anthropic-compatible API.
New · InstaChat included. Gateway + team assistant in one ₹1,000 plan — developers get the API, everyone else gets a ChatGPT-style assistant on your own keys, under the same budgets and logs. ₹1,000 + 18% GST or US$20 per 30 days, + 3% of measured net savings · 7-day free trial, no card.
- Your keysProvider accounts and invoices stay yours.
- Drop-inOpenAI & Anthropic SDKs, Claude Code.
- Made for IndiaINR, UPI, GST invoice, Aadhaar & PAN masking.
Request ledger · today₹12,480saved vs. premium-only routing
illustrativeINSTAROUTE
receipt
Premium baseline
Actual cost
You saved
prompt not stored · metadata only
₹1000
flat per month (+GST) · US$20 abroad
3%
of savings we can prove — nothing if we save nothing
29
providers you can mix behind one key
0
prompts stored — metadata only
OpenAI
Anthropic
Google Gemini
Azure OpenAI
xAI Grok
Mistral AI
DeepSeek
Groq
Cerebras
SambaNova
Together AI
Fireworks AI
DeepInfra
Nebius AI
Novita AI
FriendliAI
Baseten
Scaleway
Alibaba Qwen
Moonshot Kimi
Z.ai GLM
MiniMax
AI21 Labs
Cohere
Perplexity
OpenRouter
Cloudflare Workers AI
OpusMax
Ollama (local)
Your own endpoint
The problem with one-model AI
A greeting and a contract review shouldn't cost the same.
Most teams wire every feature to one premium model, through one provider, and find out what it cost at month-end. InstaRoute replaces that single wire with a routing layer that knows what each request needs — and keeps the receipts.
🔁
One provider, one point of failure
A 429 or an expired key stops your product. InstaRoute detours to an equivalent model on another provider automatically.
🧩
Five SDKs, five keys, five dashboards
One OpenAI- and Anthropic-compatible endpoint covers 29 providers. Change models with a setting, not a release.
🛡️
Customer data in every prompt
Emails, phone numbers, Aadhaar and PAN are masked before any provider sees them — and put back in the answer.
The InstaRoute Interchange
Four checkpoints. Three lanes. One honest bill.
Every request takes the same short drive through InstaRoute — usually a few milliseconds — and leaves on the lane its difficulty deserves. This is how the LLM router decides.
01
Guard
Personal data and secrets are masked; prompt injection is flagged or blocked.
02
Reuse
An identical or near-identical question is answered from cache, for ₹0.
03
Judge
A difficulty score decides how much model the request really needs.
04
Route
Ranked on cost, speed and quality for your chosen goal (one of 10), with detours on failure — up to 4 attempts.
05
Receipt
The response carries the model, cost, baseline and saving; Logs keep the rest.
Ask for auto, a concrete model id, or one of 10 routing goals:
- auto
- balanced
- cheapest
- fastest
- quality
- coding
- reasoning
- vision
- long_context
- local_first
Routing rules can match on goal, vision, tools, reasoning and minimum context. Failover retries on 408, 409, 429 and 5xx errors.
What you get on day one
A complete AI API gateway — nothing locked behind a tier.
⚡
A two-line switch
Point any OpenAI or Anthropic SDK at InstaRoute and keep shipping. Streaming, tools, vision and JSON mode behave exactly as before.
🔑
Your accounts, your invoices
Connect the provider keys you already pay for. InstaRoute never buys or resells tokens, so there is nothing to mark up.
⚖️
A judge for every request
Each prompt is scored for difficulty in milliseconds — easy work goes to fast, low-cost models, hard work to frontier ones.
🛣️
Detours, not downtime
A 429, an outage, an expired key or an empty credit balance triggers an instant detour to an equivalent model elsewhere.
💾
The ₹0 lane
Repeat and near-repeat questions are answered from cache. Prompt-cache-aware routing and a flex tier trim the rest.
🔒
PII masking made for India
Emails, phone numbers, cards (Luhn-checked), Aadhaar (checksum-checked) and PAN are masked before a provider sees them — and restored in the reply.
🧾
A receipt for every request
Model, attempts, tokens, cost, premium baseline and saving — in the response, in Logs and in your monthly statement.
🧯
Spend that can't run away
Budgets per key, a ceiling per workspace, and Slack or webhook alerts long before anyone gets a surprise.
🧪
Prove a model before you switch
Import your own benchmark scores and shadow-test a candidate model on a slice of real traffic.
🏠
Your own models too
Add a vLLM or any OpenAI-compatible endpoint, or Ollama for fully local inference, and route to it like any provider.
🧬
Embeddings included
14 embedding models from 7 providers behind the same key and a single /v1/embeddings endpoint.
👥
Built for teams
Owner, admin, member and viewer roles, invite links, per-app keys and an audit log of every change — up to 25 members and 180 days of request logs on Premium.
For developers
Switch in two lines. Rewrite nothing.
InstaRoute is an OpenAI-compatible API that also speaks Anthropic. Change the base URL, use an sk-inf-… key, and ask for one of 10 goals or any model by name. Every reply brings its receipt in an instaroute field.
- auto
- balanced
- cheapest
- fastest
- quality
- coding
- reasoning
- vision
- long_context
- local_first
- ✔OpenAI SDKs (Python, Node, .NET, Go, Java), LangChain, LlamaIndex, Vercel AI SDK
- ✔Anthropic Messages API — the Anthropic SDK and Claude Code
- ✔Chat Completions, Responses API (stateless), Embeddings and a free routing dry-run (/v1/route/explain, not logged)
- ✔One key per app, with its own budget, goal and provider allow-list
- ✔Rate limit: 20 requests/second per key (burst 40)
Claude Code proxy
Run Claude Code through InstaRoute.
Set one environment variable and every coding session gets failover between Claude providers, per-developer budgets, cost and savings in one dashboard, and warm prompt caches — without sharing your company's Anthropic key.
- ✔Each developer gets their own key and monthly budget
- ✔Session-sticky routing keeps the prompt cache warm
- ✔Secrets and personal data masked in prompts
- ✔Every session visible in Logs, under its own app
Terminal
# Point Claude Code at InstaRoute $ export ANTHROPIC_BASE_URL=https://instaroute.instadatahelp.com $ export ANTHROPIC_AUTH_TOKEN=sk-inf-… $ claude ✻ Welcome to Claude Code /status → base URL instaroute.instadatahelp.com ✓
New · included in Premium
InstaChat: a team AI assistant on your own keys.
InstaChat is a ChatGPT-style assistant for everyone in your workspace, at instaroute.instadatahelp.com/chat. Every message runs through the same InstaRoute routing, guardrails, budgets and logs as your API traffic — so staff chat and developer calls land in one cost report, on the provider keys you already own.
Each member gets an automatic key named “InstaChat · their email”, so spend is visible per person. Members with the Viewer role cannot use InstaChat.
Stylised illustration, not a real conversation.
Research
Web search with citations and a deep-research mode.
Work with data & files
In-browser Python code interpreter, file attachments, canvas, and local-folder editing with diff approval.
Create
Image generation and editing (needs an OpenAI key) and Sora video.
Organise
Projects, custom assistants and memory, scheduled tasks and share links.
Talk to it
Voice mode, including voice input in Indian languages.
Connect tools
Remote MCP connectors to bring your own tools into the chat.
Install it
An installable web app (PWA) for desktop and phone.
Sync privately
Optional end-to-end encrypted sync.
Some features depend on which providers you connect. Live since 8 October 2026 and documented in the InstaRoute docs.
ROI calculator
Your AI bill, rerouted.
A quick way to see how much routing could reduce your LLM costs. Once you are live, every rupee is measured per request — and the 3% only applies to what is measured.
An estimate to help you decide. Your real savings are measured on every request — the fee is based on those measurements only.
Estimated savings
InstaRoute bill
You keep
Return on the bill
Receipts, not estimates
Savings you can prove — request by request.
Each request is logged with the model that answered, what it cost, what the premium model would have cost, and the difference. Prompt text is never stored.
Overview
Spend vs the premium baseline, savings by lever, cache hit rate and latency.
Request logs
Goal, chosen model, provider, tokens, cost, savings and latency for every call.
Billing
One bill, GST shown separately, pay by UPI or card, tax invoice and receipt instantly.
Screens shown with demo data.
InstaRoute vs other AI gateways & LLM routers
Compared with Inferect, OpenRouter, Portkey and LiteLLM, using each vendor's own website.
| Capability | InstaRoute | Inferect | OpenRouter | Portkey | LiteLLM |
|---|---|---|---|---|---|
| Starting price | ₹1,000 + GST / $20 a month, all features | $17 / ₹1,199 (launch offer) | No plan fee | $49 a month + usage | Free (self-host) |
| Fee on spend or savings | 3% of measured savings | 2–3% of savings | 5.5% on credits | None | None |
| Your own provider keys, no markup | ✔ | ✔ | ✔ | ✔ | ✔ |
| Automatic per-request model choice | ✔ | ✔ | ✔ | partial | ✔ |
| Claude Code & Anthropic API | ✔ | partial | ✔ | ✔ | ✔ |
| Semantic caching | ✔ | partial | — | ✔ | ✔ |
| PII redaction with Aadhaar & PAN | ✔ | — | — | — | — |
| Savings vs baseline on every request | ✔ | partial | — | — | — |
| INR billing, UPI & GST tax invoice | ✔ | partial | — | — | — |
✔ available · partial = limited or plan-dependent · — not stated on the vendor's website. Based on each vendor's public website and documentation, checked October 2026; prices exclude taxes and change often. Product names are trademarks of their respective owners.
Simple pricing
One plan, paid from what you save.
7-day free trial · no card
InstaRoute Premium
₹1,000 + GST / month
or US$20 / month outside India
+ 3% of the savings we measure in the previous month
Start your free trial →Pay by UPI, cards, net banking or wallets (PayU) · cards via Stripe outside India · GST tax invoice with your GSTIN
Unlimited requests
all providers and models, bring your own keys, 0% markup · rate limit 20 requests/s per key (burst 40)
Every feature + InstaChat
routing, failover, caching, guardrails, statements, benchmarks, experiments, and the InstaChat team assistant
25 team members
roles, invite links, audit log, 180-day request logs
No savings, no fee
the 3% applies only to savings measured on your traffic
No lock-in
pay 30 days at a time; stop whenever you like
Local support
phone & WhatsApp +91 99724 56379, GST invoices from an Indian company
Enterprise, on request
private deployment in your own cloud, written uptime commitment, annual invoicing
Your baseline, your call
measure savings against the model you would otherwise use — set it in Billing
New · Authorised partner programme
Buy locally — or become an authorised partner.
Since 11 October 2026, InstaRoute can be sold and supported by authorised partners. Each partner gets a co-branded landing page, and customers who arrive through it are attributed to that partner. Partner terms are part of the InstaRoute Terms of Service.
instaroute.instadatahelp.com/partner/<your-code>
Co-branded landing page
Your own /partner/<code> page showing InstaRoute “with your authorised partner”.
30-day attribution
Sign-ups that come through your link are credited to you for 30 days.
Your customers only
A scoped partner admin view: you see and manage only the customers you brought in.
Trials & payments
Grant free access of up to 30 days per action and record customer payments.
Commission reports
Track your commission, with QR and WhatsApp sharing for your link.
For customers
Prefer to buy through a local reseller? Ask us whether an authorised partner can serve you.
Frequently asked questions
What is an AI gateway or LLM router?+
An AI gateway sits between your application and AI providers such as OpenAI, Anthropic and Google. An LLM router decides which model should answer each request. InstaRoute does both: one OpenAI-compatible API in front of 29 providers that picks the best-value model for every request, fails over automatically and records the cost and savings.
How does InstaRoute reduce LLM and OpenAI API costs?+
Most requests do not need the most expensive model. InstaRoute scores each request for difficulty and sends simple ones to fast, low-cost models and only hard ones to premium models. Exact and semantic caching, prompt-cache-aware routing and a flex tier add further savings. Every response shows the cost against the premium baseline, so the savings are measured, not estimated.
Is InstaRoute compatible with the OpenAI API and SDKs?+
Yes. Change the base URL to https://instaroute.instadatahelp.com/v1 and use your InstaRoute key — the OpenAI SDKs for Python, Node, .NET, Go and Java, LangChain, LlamaIndex and the Vercel AI SDK all work. Streaming, tool calling, vision and JSON mode are supported. The Responses API is supported in stateless mode only (previous_response_id, conversation, background and built-in tools are not available).
Can I use Claude Code with InstaRoute?+
Yes. InstaRoute speaks the Anthropic Messages API. Set ANTHROPIC_BASE_URL to https://instaroute.instadatahelp.com and ANTHROPIC_AUTH_TOKEN to your InstaRoute key, then run claude. Each session gets failover, cost tracking and per-key budgets.
Do I pay InstaRoute for tokens?+
No. You bring your own provider keys and pay providers directly at their prices — InstaRoute never marks up tokens. You pay InstaRoute ₹1,000 + GST a month (US$20 outside India) plus 3% of the savings it measures.
Which AI providers and models are supported?+
29 providers including OpenAI, Anthropic, Google Gemini, Azure OpenAI, xAI Grok, Mistral, DeepSeek, Groq, Cerebras, Together, Fireworks, OpenRouter, Perplexity, Cohere, Alibaba Qwen, Moonshot Kimi and Z.ai GLM, plus Ollama and any OpenAI-compatible endpoint you host — 140+ routable models, and hundreds more through OpenRouter.
What routing goals can I choose, and are there rate limits?+
Set the model field to a concrete model id, or to one of 10 goals: auto, balanced, cheapest, fastest, quality, coding, reasoning, vision, long_context and local_first. Routing rules can also match on goal, vision, tools, reasoning and minimum context. Each key is limited to 20 requests per second (burst 40), and a failed call is retried on another model for up to 4 attempts.
What is InstaChat and does it cost extra?+
InstaChat is a ChatGPT-style team assistant at instaroute.instadatahelp.com/chat, included in the Premium plan at no extra plan fee — you pay your AI providers for what it uses, as with any other request. It offers web search with citations, deep research, an in-browser Python code interpreter, file attachments, canvas, projects, assistants and memory, voice mode with Indian-language voice input, remote MCP connectors, scheduled tasks and share links. Every message runs through your workspace routing, guardrails, budgets and logs. Members with the Viewer role cannot use it.
Can I resell InstaRoute or buy through a local partner?+
Yes. The authorised partner programme went live on 11 October 2026. Partners get a co-branded landing page at instaroute.instadatahelp.com/partner/ followed by their code, 30-day attribution for sign-ups, a scoped admin view of their own customers and commission reports. Contact us to apply, or to ask whether a partner can serve you.
Is my data safe? Are prompts stored?+
Prompt and reply text are never written to logs or the database — only metadata such as model, tokens and cost. Provider keys are encrypted at rest (AES-256-GCM), and guardrails can mask emails, phone numbers, card, Aadhaar and PAN numbers before any provider sees them.
How are savings measured — and can I set my own baseline?+
Each routed request is compared with a baseline: by default the most expensive model that could have done the job. You can instead declare the model you would otherwise use (for example GPT-4o) under Billing, and savings are then measured against that. If a request needs a pricier model, the difference counts against the saving. The baseline used is stored with every request and named on your monthly statement.
Can InstaRoute run inside our own cloud?+
Yes, on request. Enterprise customers can run a private deployment of the full gateway in their own VPC or data centre, with a written uptime commitment, dedicated onboarding and annual invoicing. Live service health is public at instaroute.instadatahelp.com/status.
How does billing work in India?+
After the 7-day free trial you get one bill every 30 days: ₹1,000 plus 3% of the previous period's measured savings, with 18% GST shown separately. Pay by UPI, card or net banking through PayU and receive a GST tax invoice with your GSTIN instantly. There is no auto-debit and no contract.
Your first receipt is five minutes away
Connect one provider key, change one base URL, send one request — and open InstaChat for the rest of your team. Every feature, free for 7 days — no card needed.
instaroute.instadatahelp.com · a product of InstaDataHelp AI Services Private Limited
