Run analytics, indicators and backtests where useful CPU and memory already exist.
Keep analytics, indicators and risk checks in Colab, Jupyter or your own machine. Route only approved create, modify and cancel calls through a dedicated IPv6 your broker can recognize.
Use the numeric client ID shown in DhanHQ · No card. Broker secrets stay in your runtime.
Recommended algo workflow
Heavy pandas, Jupyter, backtests and multiple data feeds can exhaust a 1 GB VM or saturate shared CPU until the machine becomes slow or unreachable. Free and low-cost VPS plans may also throttle sustained burst usage or send a fair-use notice under their own policies.
Run analytics, indicators and backtests where useful CPU and memory already exist.
trading-static-ipApproved broker calls leave through your assigned ServLoci IPv6; unrelated notebook traffic stays direct.
StaticIP supplies the stable broker-facing path. Use managed compute for unattended or continuously running strategies.
Colab is interactive research compute, not a 24×7 trading host. Runtimes can disconnect and resource limits vary. For unattended execution, keep the same ServLoci exit IP but move the strategy to managed, persistent compute. Google Colab runtime FAQ ↗
The Problem
You didn't do anything wrong. This is how residential and mobile networks work in India — and how brokers verify who's allowed to call their API. The mismatch is the trap.
Jio, Airtel, ACT — most Indian ISPs hand out dynamic addresses. A router reboot, a link flap, an ISP-side lease renewal, and your public IP silently changes. Every broker whitelisted to the old one blocks you instantly.
Many residential and mobile connections use CGNAT, where multiple subscribers share one public IPv4. That address is a poor fit for an account-specific broker allowlist.
A hosted notebook can move to a different runtime when it reconnects. Your research still works, but the outbound address may no longer match the one approved for broker order routing.
Started a session in the morning at IP 203.0.113.42. Router disconnected at 11am. Reconnected at 45.117.8.99. The broker still holds an open session tied to the old IP — and refuses your new calls until the token expires. Locked out of your own account.
4G/5G on Jio, Airtel, Vi — carriers rotate your public IP every few minutes as you move between towers. Your algo running off a mobile hotspot survives about one order cycle before the address changes. No whitelist can keep up.
Managing algos for yourself + family + partners across Dhan + Kite + Groww? To keep whitelist collisions from taking everyone down at once, you spin up a VPS per account. Costs stack: ₹500-800/mo per account, plus the ops time to babysit them.
Dead Ends
A VPS can provide a fixed address, but it also makes you operate a server just to satisfy a networking requirement. Multiple identities can multiply infrastructure and maintenance work.
Broker-hosted tools can reduce setup for that broker, but they may not fit a cross-broker workflow or the compute environment where your strategy already runs.
Managed algo platforms can accelerate standard strategies, but custom workflows must fit their supported features, integrations, limits and pricing.
Shared exit IPs can inherit the traffic reputation and rate behaviour of unrelated customers. They also cannot provide the account-specific network identity a dedicated allocation offers.
The math
When the source address no longer matches an approved route, execution can stop until the network or broker configuration is corrected. The impact depends on the strategy and broker.
A strategy can produce a valid signal while the broker rejects the request because the observed source no longer matches the approved route.
A changed source address can mean updating settings or contacting broker support, depending on that broker's current process.
A VPS can provide a fixed IP, but it also adds patching, monitoring, deployment and credential-management work when the compute itself was not needed.
A residential or pooled address may represent unrelated users, which makes it a weaker identity for an account-specific broker allowlist.
The Solution
One dedicated IPv6 address, yours alone. Route the broker-facing calls that require an approved source through it via SOCKS5 or the restricted HTTP forwarder.
What you actually get
The dedicated IPv6 service is live today. A unified trading library and trade-management layer are separate private-beta products in development.
One dedicated address for the life of the active allocation. Register it with compatible broker APIs, then use the same broker-facing exit from your notebook, workstation or automation runner.
One Python (and JS) SDK that speaks Dhan, Kite, Groww and Fyers. Same method names. Same order shape. Same tick-stream interface. Add a new broker by changing one string — not rewriting a service.
tradeclient.place_order(broker="kite", ...)A planned dashboard for supported accounts, routing rules, fills, exposure and P&L. Private-beta scope may change as broker integrations are validated.
Paying subscribers can request early beta access. Beta scope, availability and future pricing may change.
Under the hood
A static exit IP is a solved, boring networking problem. We've built the plumbing so you don't have to think about it.
A /128 is the smallest IPv6 allocation — a single address. Assigned to you alone. Not a range shared with other users, not a pool that rotates. One address. Yours.
Every connection requires your username and password. Those credentials map to exactly one IPv6 exit. No auth = no connection. There's no anonymous routing, no shared session.
With SOCKS5, broker HTTPS remains TLS-encrypted between your client and the broker. The optional /api/v1 application forwarder is a separate path that processes customer-authorized order requests.
The IPv6 is configured on the server's network interface — not a floating IP, not a NAT rule. It doesn't change when software restarts, when we deploy updates, or when you reconnect.
Python's requests library supports SOCKS5 via PySocks. Monkey-patching the session before your broker import means the SDK never knows it's proxied. No broker SDK modification.
We give you a stable broker-facing exit IP. Your strategy and execution logic stay on your compute. Choose SOCKS5 for broker end-to-end TLS, or the restricted /api/v1 forwarder for authorized order and trade paths.
Setup
Google sign-in. Share your broker name and client ID. We assign a dedicated IPv6 and SOCKS5 credentials.
~2 minutesThree lines before your broker SDK import. No VPS config, no nginx, no iptables. The SDK never knows it's proxied.
3 lines of PythonAdd your IPv6 to each broker's whitelist settings. Done. Forever. Move machines as often as you like — the exit IP never changes.
dedicated while activeAdvisory
Buy one address, run it lean. Six habits the sharpest ServLoci users hold — each one keeps your monthly bill flat while the strategy behind it gets fatter.
The same IPv6 can be used with an approved account on Dhan, Kite, Groww, FYERS, Upstox, ICICI Direct or Kotak Neo. Broker approval and IP-family rules still apply. Don't buy a second address until you actually need a second identity.
10 GB / month is included. If a busy month sneaks past that, overage bills at ₹5 per extra GB — no upsell, no plan switch. Most order-only workloads run under 2 GB. If you're at 30, you're probably routing traffic that doesn't need to be routed.
Only create, modify, cancel need to leave from your whitelisted IP. Ticks, LTP snapshots, holdings, positions, historical bars — those can go straight from your box. Same account, same broker; nothing breaks. Data usage on our side drops by an order of magnitude.
A ₹800/mo VPS gives you 2 vCPUs and a rotating IP. Your desk already has a faster machine that isn't rotating anywhere. Run the strategy there — the proxy is what makes the exit address stable, not the VPS. Ops shrinks to "did I close the laptop lid?"
Backtests, feature engineering, model inference, indicator computation — none of it hits the network. Only the final order dispatch touches ServLoci. Your compute stays with you, your data stays with you, your competitive edge stays with you.
An agent watching signals, computing sizing, punching the trade — it all lives on your box. The single line of egress config is what makes it deployable. Order came from anywhere in the world? Broker sees the same IPv6. The agent doesn't care where it runs; you don't care what changed.
When to add a second IP
One address per trading identity, not per broker. A family account, a partner book, a bot that needs its own whitelist — that's when you provision a second IP. Everything else stays on the one you already pay for.
Integration
Configure socks5h before importing your broker library. The proxy accepts only your selected broker's approved HTTPS API destinations.
Pricing
Use our hosted IPv6, add IP Manager to a VM you control, or choose the fully managed opaque appliance. No setup fee. Cancel anytime.
Explorer is free with no card. Upgrade to ₹300/month to retain the same allocation after day seven.
VM charges and OS administration remain yours. This is not the fully managed opaque appliance.
Approximate list price. Opaque appliance-style install — not open-source DIY. Contact us to size the host.
Already have AWS, Google Cloud, Azure, Oracle Cloud, or Vultr credits? Use our provider-neutral installer to add Python and all seven broker SDKs to an Ubuntu/Debian VM you operate. Open the free VPS installer →
FAQ
/api/v1 forwarder operates at the application layer and processes only customer-authorized order and trade paths. Broker secrets used for verification are not persisted.Start today
Sign in with Google → paste your broker + client id → we assign a dedicated IPv6 with 1 GB / 10 RPS / 7 days on the house. No card. Upgrade later — the same IP survives.
Claim my free IPv6 · 1 GB · 7 days · no card Read the docs firstor submit your broker details directly