India's sovereign
AI inference cloud
IndieRouter is an OpenAI-compatible API for chat, coding and agents on open-weight models, served from our own GPUs in India, never routed to offshore inference APIs. Pay as you go, or monthly plans from ₹499$5, with GST invoices.
Signing in with Google or GitHub shares only your name and email address, used solely to create and secure your IndieRouter account. Details in our Privacy Policy.
OpenAI-compatible · Prepaid credits · Prompts never retained or trained on
client = OpenAI(
- base_url="https://api.openai.com/v1",
+ base_url="https://api.indierouter.ai/v1",
api_key=os.environ["INDIEROUTER_API_KEY"]
)
r = client.chat.completions.create(
model="deepseek-v4.1-flash", # the model itself, no aliases
messages=[{"role": "user", "content": "Namaste, world"}],
stream=True
)How requests flow
Your app to Indian GPUs. Nothing in between.
Every request terminates inside the country. Failover stays in-country, never region-to-abroad.
Failover stays in-country, never region-to-abroad
Quickstart
Migrate in one line
IndieRouter implements the OpenAI API surface. Your SDK, streaming and function calling carry over unchanged: only the base URL moves.
Not real-time? POST /v1/batchesruns the same surface as an async job at a discounted rate: upload a JSONL, poll or get a signed webhook. Same zero retention, same India-only compute.
# the entire migration from OpenAI:
client = OpenAI(
base_url="https://api.indierouter.ai/v1",
api_key=os.environ["INDIEROUTER_API_KEY"],
)
r = client.chat.completions.create(
model="deepseek-v4.1-flash",
messages=[{"role": "user", "content": "Namaste, world"}],
)Console
Not just an API. A place to build.
Every account gets a full chat workspace in the browser: try any model, give it your own tools, and let it write, run and preview real code, without leaving the page.
Chat with every model
Streaming replies, visible reasoning, and image input on vision models. Switch models mid-thread; your history stays in your browser.
Bring your own tools (MCP)
Attach MCP servers and the model calls your tools inline: your data, your APIs, your actions, wired into the conversation.
Workspaces that build apps
Ask for an app or a site and the model gets an isolated Linux workspace to write and run code, with a live preview link and a one-click project download.
Web search and browsing
The model can search the open web and read live pages when a question needs current facts, then cite what it used.
Models & pricing
Every model, direct. No route aliases.
The model parameter takes the model itself: you pick the exact open-weight model, we publish its card, guarantee the SLA and residency, and announce version changes 60 days ahead. Model origin is not hosting location: every model serves from Indian GPUs.
The economics of residency
Priced for sovereignty, not a race to the bottom
Offshore token discounters can undercut on sticker price, and cannot legally hold Indian regulated data. DeepSeek's own API bills its peak rates, about 2x off-peak, exactly during Indian working hours; our rate card is flat 24/7. What you buy here is residency, plus the costs that disappear when the invoice is Indian.
Data residency, guaranteed by design
Most providers sell residency as an addendum. We have no abroad.
IndieRouter cannot send your data abroad: there is no abroad in the architecture. Routing policy is part of the product.
Every byte stays in India
Prompts, files, logs and tool outputs are processed and stored exclusively in Indian data centres: Indian storage, Indian backups, Indian support paths.
Zero-retention inference
Request bodies are discarded once the response is served. We meter tokens, not content: nothing to leak, nothing to train on.
No offshore routing, ever
We use the best open-weight models regardless of origin, but hosting is only ever India. No request touches a China-hosted or offshore inference API.
Built for regulated India
Built for India's regulated sectors
We are actively pursuing government, BFSI and healthcare deployments, the sectors that legally cannot export data. Designed to meet their bar; certifications named are targets we are working toward, not claims.
Government & PSU
Forms, circulars and multilingual citizen-service copilots under data-localisation mandates. Pursuing MeitY empanelment and a GeM listing so departments can procure through existing rails, with air-gapped single-tenant options.
Banking & insurance
Policy search, audit Q&A and call-transcript summarisation inside the RBI storage-in-India expectation and IRDAI/SEBI norms. Zero retention means customer text, statements and claims are never stored or trainable.
Healthcare & life sciences
Aligned to ABDM / ABHA data handling and DPDP sensitive-personal-data processing on Indian soil. De-identification-friendly, zero-retention pipelines for clinical documentation, coding and triage.
Engagement model: a scoped, dedicated-endpoint pilot with a signed DPA, then expand. Ask for our security & compliance one-pager.
Enterprise trust
What BFSI and Government actually ask for
Data controls
India-only inference, no training on customer data, retention controls, PII redaction, per-tenant encryption, audit export.
Network controls
Private endpoint, IP allowlist, VPC peering or tunnel, dedicated capacity, separate logging namespace.
Procurement controls
Model provenance sheet, open-weight licence review, DPDP-aware DPA, SLA, support escalation, GST invoice.Model provenance sheet, open-weight licence review, DPDP-aware DPA, SLA, support escalation.
Operational controls
p95 latency SLO, status page, incident reports, model version change notice, abuse monitoring.
AI controls
Prompt-injection guardrails, model denylist, output filters, eval reports, hallucination and retrieval scoring.
Commercial controls
Committed spend, reserved capacity, fair-use queues, minimum contract size for long-context lanes.
Dedicated endpoints
Dedicated capacity, fully managed
Reserved capacity with throughput SLOs, network isolation, audit logs and a signed DPA: for BFSI, healthcare and government workloads that need more than a promise.