An NPCI initiative
AiNxt Suite — from NPCI

One Intelligence.
Works for Everyone.

AiNxt is a suite of AI products designed to bring intelligence into enterprise work, development environments, command-line workflows, and the foundations used to build new AI experiences.

Four products. All open source. Yours to use and build on.
Explore the Suite What is AiNxt? Open-source repositories ↗
01AiNxt Enterprise 02AiNxt OS 03AiNxt Code 04AiNxt CLI
The idea

What is AiNxt?

AiNxt brings AI into the places where people and organisations work — from enterprise applications and everyday work, to software development and the platforms used to build new AI capabilities.

For individuals

Use intelligence where you already work: ask questions, understand information, create and modify software, automate tasks, work with agents, and interact with AI from the browser, desktop, IDE or terminal.

For organisations

Bring AI into teams and enterprise workflows with a focus on governed access, organisational context, security, accountability and the ability to build capabilities tailored to business needs.

AiNxt Suite

Four products. One suite.

AiNxt Suite brings together four distinct products, each designed for a different way of working with AI.

01

AiNxt Enterprise

The governed enterprise AI environment, across web and desktop.
02

AiNxt OS

The foundation for building your own applications, agents and workflows.
03

AiNxt Code

AI inside the editor — complete, rewrite, explain and fix.
04

AiNxt CLI

AI in the terminal — ask, build, fix, automate and execute.
ProductWhat it providesPrimary users
AiNxt EnterpriseEnterprise AI experiences across web and desktop, including conversational, agentic and software-lifecycle capabilities.Individuals, teams and organisations
AiNxt OSFoundational capabilities for building and running AI applications, agents, workflows and custom experiences.Developers, teams and organisations
AiNxt CodeAI-assisted software development inside the IDE.Developers
AiNxt CLIAI-assisted work and execution from the terminal.Developers and technical teams
See how the whole suite fits together
How it all fits together

One suite. Five layers.
Everything governed.

AiNxt is not four separate tools that happen to share a name. It is one intelligence layer that shows up differently depending on where you work — and every one of those surfaces stands on the same foundation, passes through the same guardrails and draws on the same choice of models. Read the diagram from the top down: the people it serves, the places it appears, the four products they actually use, the trust spine that governs every single interaction, and the models underneath it all.

Scope: the whole AiNxt Suite. All four products appear as equals in layer 03 — none of them is a satellite of another.

01

People

Who AiNxt is built for — not a pilot group, the whole organisation
Individuals · Teams · Enterprise
Every employee Ask, understand, draft and decide with intelligence that already carries the enterprise context — no prompt engineering required.
Business teams Turn recurring departmental work into agents and workflows that run the same way every time.
Developers Stay in the editor and the terminal. Intelligence arrives where the code already is.
Platform & technical teams Build new AI products on a foundation that is already governed, observable and model-agnostic.
Risk, security & leadership See what AI is doing across the organisation — usage, quality, cost and behaviour — in one place.
meet intelligence at
02

Surfaces

Where the intelligence appears — you go to your work, not to a chatbot
Five ways in, one brain behind them
Browser The full AiNxt Enterprise experience, nothing to install, available to every employee.
Desktop A dedicated app that reaches the local machine and the files and tools that live on it.
IDE AiNxt Code inside the editor — completion, rewrite, explain and fix, next to the cursor.
Terminal AiNxt CLI in the shell developers already live in, scriptable and automatable.
Office Add-ins that put AiNxt inside the documents, mail and decks where work actually gets written.
delivered as four products
03

The AiNxt Suite

Four products, each aimed at a genuinely different way of working with AI
The layer people actually touch
01Web + Desktop

AiNxt Enterprise

The governed enterprise AI environment. Everyday work, organisational knowledge, agents, the software lifecycle and the local desktop — in one experience.

  • AiNxt ChatConversational access to enterprise-aware intelligence for questions, analysis and daily work.
  • AiNxt BuddyA coworker, not a chat window — works the local machine, connects to Gmail, Outlook and approved tools, and takes your own persona.
  • Enterprise Knowledge HubOrganisational knowledge gathered in one place so people can discover it, understand it in context and act on it.
  • AiNxt CoachAn AI mentor that teaches every employee to actually use AI well, and keeps improving how they work.
  • Agent StudioBuild specialised agents around real business needs, tasks and workflows.
  • ADLCAgentic Development Lifecycle — intelligence applied across the enterprise software lifecycle.
Explore Enterprise
02Foundation

AiNxt OS

The foundation everything else is built on — and the one you build your own products on. The experience does not have to look like AiNxt.

  • ApplicationsBuild AI-powered applications around your own use cases, on infrastructure that is already governed.
  • AgentsCreate specialised intelligence that performs defined work, with defined tools and defined limits.
  • WorkflowsConnect intelligence, tools and business processes into repeatable, reviewable flows.
  • Tools & connectorsGive your agents governed access to enterprise systems without handling credentials yourself.
  • Eval & telemetryMeasure what you build with the same evaluation and observability the products use.
  • Your own facePut your own interface on top. AiNxt OS can power purpose-built products for entirely different domains and users.
Explore OS
03IDE plugin

AiNxt Code

Intelligence inside the developer workspace, for understanding, creating, modifying and reviewing software without leaving the file.

  • Code completionContextual suggestions as you type, aware of the code around them.
  • Rewrite & refactorTransform a selection with AI assistance instead of retyping it.
  • ExplainUnderstand unfamiliar code and the implementation decisions behind it.
  • FixIdentify and address coding issues with assistance, in place.
  • Developer contextIntelligence stays close to the code actually being worked on.
  • Governed accessDeveloper AI usage is connected to the same enterprise controls as everything else.
Explore Code
04Command line

AiNxt CLI

Intelligence in the terminal, so technical teams can ask, build, troubleshoot, automate and execute without leaving their existing workflow.

  • AskInteract with intelligence without opening a browser.
  • BuildAI assistance through terminal-driven development tasks.
  • FixInvestigate and troubleshoot technical problems with assistance.
  • AutomateBring intelligence into repeatable command-line workflows and scripts.
  • ExecuteMove from an instruction toward an actual action in the terminal.
  • GovernTerminal AI usage stays inside organisational controls.
Explore CLI
every request passes through
04

The trust spine

Not a feature of one product — a layer every product is required to pass through
Applies to all four products, on every surface
Built-in guardrails Defence against jailbreak attempts, prompt injection, prompt poisoning and related prompt-based attacks — part of the experience, not something each employee has to manage.
Eval Evaluate AI behaviour and answer quality deliberately, so changes are measured rather than assumed.
Observability Understand usage, performance, quality and outcomes across the organisation — one view, not four.
Governance & access Organisational controls and accountability follow the user, whichever product or surface they came in through.

This is the layer that makes the difference between an AI demo and an AI platform an enterprise can actually stand behind.

and runs on the models you choose
05

Models — BYOM

Bring Your Own Model: the suite is designed around model choice, not model lock-in
Swap the model, keep the platform
Hosted by the provider Connect approved models from Anthropic, OpenAI and Google over their APIs, for work that needs the strongest available reasoning.
Open-weight, self-hosted Run Kimi, DeepSeek, Qwen and other approved open-weight models on your own infrastructure, so nothing leaves it.
Chosen per use case Route each use case on capability, performance, privacy, deployment requirements and cost — without rewriting the product above it.
Learn it once The same intelligence, the same context and the same controls whether you opened a browser, an IDE or a terminal.
Governed by default Guardrails, evaluation, observability and access control are not add-ons you remember to switch on. They are the path every request takes.
Yours to build on Use the four products, or take AiNxt OS and build something that looks nothing like them. Both are first-class.
Now follow a single request through it
Follow one request

From a question
to a governed answer.

The layer diagram above shows what AiNxt is. This one shows what actually happens when somebody uses it. A single request — typed in a browser, an editor or a terminal — travels down through seven stages and comes back up. Nothing skips a stage: the same identity check, the same guardrails and the same observability apply whether the request came from AiNxt Chat or a script in a terminal.

Scope: the whole AiNxt Suite. Stage 01 is every client — browser, desktop, IDE, terminal and Office — and stages 02 to 07 are identical for all of them.

forward — request, context, tool call return — answer, citations, telemetry
01

Clients

Web appbrowser · full Enterprise experience
Desktoplocal machine · local files and tools
IDEAiNxt Code · inline in the editor
CLIAiNxt CLI · scriptable, automatable
Office add-insdocuments · mail · decks
request + user identity streamed answer
02

Gateway

API gatewayone entry point for every client
Sessions & streamingtokens streamed back as they are produced
Limits & quotasrate limits · back-pressure · fair use
authenticated call authorised response
03

Identity
& access

Enterprise SSOthe organisation's existing directory, not a second login
Role-based accesswhat this person may see, ask and do
Per-user scopingevery query narrowed before it runs, never filtered after
scoped to what the user may see permitted content only
04

Guardrails

Prompt inspectionevery prompt sanitised before it reaches a model
Injection & jailbreak defenceprompt injection · poisoning · adversarial input
Redact and proceedsensitive data is removed, the user is never dead-ended
Output compliancethe answer is checked on the way back out, too
safe, in-policy prompt checked, compliant output
05

Orchestration

Chatthe conversational path · intent, context, answer
AgentsAgent Studio · specialised work with defined tools
Workflows & ADLCrepeatable, reviewable multi-step work
Tool sandboxtools run isolated · credentials never handed to the model
retrieve · read · write grounded context + citations
06

Knowledge
& retrieval

Knowledge Hubenterprise documents and organisational knowledge
Hybrid searchmeaning and keywords, ranked together
Code & repo indexthe codebase, parsed and searchable
Knowledge graphhow documents, code and people relate
prompt + grounded context completion
07

Models

Model routerpicks the model per use case, not per product
Hosted by the providerAnthropic · OpenAI · Google, over the API
Open-weight, self-hostedKimi · DeepSeek · Qwen, on your infrastructure

Read the two columns of arrows. The dark arrows going down are the request being narrowed — identified, scoped, sanitised, grounded. The green arrows coming up are the answer being checked on its way back. A request cannot reach a model without passing stages 03 and 04, and an answer cannot reach a person without passing stage 04 again.

01 • Enterprise
Web + Desktop

AiNxt Enterprise

A complete enterprise AI environment that brings intelligence into everyday work, enterprise knowledge, agents, development workflows and the local desktop — with model choice, observability and built-in protection across the experience.

Enterprise AI suite
For an individual

Work with AI that understands your enterprise context, helps you find and use organisational knowledge, acts as a coworker across local and connected tools, supports development and routine work, and helps you become more effective with intelligence.

For an organisation

Give employees a governed enterprise AI environment with enterprise knowledge, agents, model choice, observability, security guardrails and experiences across web and desktop.

Core enterprise capabilities

AiNxt Chat Conversational access to enterprise-aware intelligence for questions, analysis and everyday work.
AiNxt Buddy A coworker-like assistant that works on the local machine, connects to external apps and helps organise work and routine jobs — with the ability to add your own persona.
Enterprise Knowledge Hub Bring enterprise knowledge together in one place so people can discover, understand and use organisational information.
AiNxt Coach An AI mentor inside the enterprise that helps individuals learn effective AI usage and improve productivity.
Agent Studio Create specialised agents around business needs, tasks and workflows.
Office Add-ins Bring AiNxt intelligence into everyday Office workflows where people already work.
ADLC Agentic Development Lifecycle — apply intelligence across the enterprise software development lifecycle.
Eval & Observability Evaluate AI behaviour and provide an observability layer for understanding quality, usage, performance and outcomes.
Built-in Guardrails Defence mechanisms for threats such as jailbreaks, prompt injection, prompt poisoning and unsafe AI interactions.
Bring Your Own Model

BYOM — use the models that work for you

AiNxt Enterprise is designed to work across model providers rather than locking the enterprise into a single LLM. Organisations can connect approved models from providers such as Anthropic, OpenAI and Google, as well as open-weight models they host themselves — such as Kimi, DeepSeek and Qwen — on their own infrastructure.

This gives organisations flexibility to choose models based on capability, performance, privacy, deployment requirements and use case.

Desktop experience

AiNxt Enterprise also includes a dedicated desktop experience, bringing enterprise intelligence into the user's local working environment. This complements the web experience so intelligence can be available where work actually happens.

One environment, many ways in

Enterprise is not a single app but a set of experiences over one shared spine. Whichever one you open, it draws on the same organisational context and passes the same controls.

01

Where you work

Browserthe full experience, nothing to install
Desktopreaches the local machine and its files
Office add-insinside the documents and mail themselves
you open one answer, in context
02

The experiences

AiNxt Chatask, analyse, decide
AiNxt Buddya coworker that acts, not just answers
Knowledge Huborganisational knowledge, discoverable
AiNxt Coachlearns you, and teaches you
Agent Studiobuild agents for your own work
ADLCintelligence across the software lifecycle
asks for context grounded, cited answer
03

Enterprise context & control

Organisational knowledgedocuments, systems and history
Your rolescoped to what you specifically may see
Guardrailsinjection defence and compliance, every request
Eval & observabilityquality, usage and cost, in one view
prompt + grounded context checked completion
04

Models — BYOM

Hosted by the providerAnthropic · OpenAI · Google, over the API
Open-weight, self-hostedKimi · DeepSeek, on your own GPUs
Chosen per use casecapability, privacy, deployment, cost
Explore the Suite GitHub repository Developer resources
02 • Foundation
Build your own AI experiences

AiNxt OS

A foundational AI platform for developers and organisations to build their own AI applications, agents, workflows and experiences.

Foundation
For an individual

Build and customise AI experiences for the way you work — from specialised assistants and agents to workflows that automate recurring tasks.

For an organisation

Give teams a common foundation for creating AI capabilities tailored to their domains, processes and business needs.

Build your own face on top of AiNxt OS

AiNxt OS is intended as a foundation that developers and organisations can build their own interfaces, applications, agents and experiences on top of. The experience does not have to look like AiNxt Enterprise; the foundation can power purpose-built products for different domains and users.

Where your code stops and the platform starts

AiNxt OS is the only product in the suite you build on rather than use. The dashed layer is yours; everything below it you inherit.

01

You build

Your interfaceweb, desktop, mobile or headless — it does not have to look like AiNxt
Your domain logicthe rules and vocabulary of your business
Your datayour systems of record stay yours
you compose primitives
02

OS primitives

Agentsspecialised intelligence with defined tools and defined limits
Workflowsrepeatable multi-step work you can review
Tools & connectorsgoverned access to enterprise systems
Memory & contextwhat an agent knows, and for how long
Evalmeasure behaviour instead of assuming it
runs on governed by default
03

Inherited platform

Identity & accessthe same SSO and role-based scoping as every product
Guardrailsinjection defence and compliance you do not implement yourself
Observabilitytraces, quality signals, latency and cost
Retrievalhybrid search over enterprise knowledge and code
prompt + context completion
04

Models — BYOM

Model routerchoose per use case, swap without rewriting
Hosted by the providerAnthropic · OpenAI · Google, over the API
Open-weight, self-hostedKimi · DeepSeek, on your own GPUs
GitHub repository Documentation Learn more
03 • Developer
Intelligence inside the IDE

AiNxt Code

Bring AI directly into the developer workspace for understanding, creating, modifying and reviewing software.

IDE plugin
settlement.py Ghost text
15@router.post("/settlement/batch")
16async def settle_batch(req: BatchRequest):
17 txns = await repo.fetch_pending(req.batch_id)
18 for txn in txns:
19 await ledger.post(txn, ref=req.batch_id)
Tab to accept
Inline completion. The suggestion on lines 18–19 is not in the file yet — it is dimmed until you press Tab. Sent from the buffer around your cursor, and served by a low-latency model so it keeps up with typing.
settlement.py Select → rewrite
17- txns = await repo.fetch_pending(req.batch_id)
17+ txns = await repo.fetch_pending(
18+ req.batch_id, limit=BATCH_MAX
19+ )
20+ if not txns:
21+ return SettlementResult.empty(req.batch_id)
Accept hunkRejecthunk 1 of 2
Select → rewrite. Highlight a block, describe the change, and the replacement streams back as an inline diff. You accept or reject it hunk by hunk — the rewrite never lands as one all-or-nothing edit. Representations of the behaviour specified in the IDE product spec — not captured screenshots.
For an individual

Get suggestions as you work, ask for code to be explained or rewritten, fix issues and work with AI without constantly leaving the file or editor.

For an organisation

Give developers an AI-assisted development experience while retaining organisation-level governance and controls around AI access and usage.

Code completionGet contextual suggestions while writing code.
Rewrite & refactorTransform selected code with AI assistance.
ExplainUnderstand unfamiliar code and implementation details.
FixUse AI assistance to identify and address coding issues.
Developer contextKeep intelligence close to the code being worked on.
Governed accessConnect developer AI usage to enterprise controls.

From cursor to change

AiNxt Code turns what is already on screen into the context for a suggestion — then sends it down the same governed path as every other surface.

01

In the editor

Cursor contextthe lines you are writing, right now
Selectionthe block you highlighted to rewrite or explain
Open fileswhat else you have in front of you
Project indexthe wider codebase, parsed and searchable
becomes context suggestion at the cursor
02

What it does

Completecontextual suggestions inline as you type
Rewritetransform a selection instead of retyping it
Explainunderstand unfamiliar code and the decisions behind it
Fixidentify and address issues in place
every request checked output
03

The same governed path

Identity & accessyour organisational role follows you into the editor
Guardrailsthe same inspection and compliance as the browser
Observabilitydeveloper AI usage is visible alongside everything else
prompt + code context completion
04

Models

Low-latency modelfor inline completion, where milliseconds show
Reasoning modelfor rewrite, explain and fix
Governed choicethe organisation decides which models are approved
GitHub repository Documentation
04 • Terminal
Intelligence from the command line

AiNxt CLI

Bring AiNxt intelligence into the terminal so developers and technical teams can ask, build, troubleshoot, automate and execute without leaving their existing command-line workflow.

Command line
ainxt AiNxt CLI
feat/idempotent-settlement ~/npci/payments-api
❯ make the settlement endpoint idempotent
⏺ Reading the settlement endpoint and its tests.
Searched 3 patterns, read 2 files (ctrl+o to expand)
└ routers/settlement.py
└ tests/test_settlement.py
⏺ Proposing a change. Nothing has been written yet.
╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌
Plan make the settlement endpoint idempotent
1. Add an Idempotency-Key check to routers/settlement.py
2. Wrap the batch call in ensure_once() so a replay is a no-op
3. Cover the replay case in tests/test_settlement.py
╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌
Proceed with this plan?
Yes, apply ity
Edit the plane
No, tell ainxt what to changen
plan mode · esc to interruptainxt 0.2.101 Beta
It proposes, you confirm, then it acts. Plan mode is on by default, so the agent states what it intends to do and stops. Nothing reaches the working tree until you approve — --always-approve exists precisely because approval is the default. For CI, the agent subcommand runs the same binary headlessly. Representation of documented behaviour, drawn to the real TUI's conventions — not a captured session.
For an individual

Stay in the terminal you already use. Ask for a change, a fix or an explanation, and use AI-assisted capabilities directly from the command line.

For an organisation

Extend AI into technical workflows while retaining the same expectation of governed access, compliance controls and accountability.

AskInteract with intelligence without opening a browser.
BuildUse AI assistance while working through terminal-driven development tasks.
FixInvestigate and troubleshoot technical problems with AI assistance.
AutomateBring intelligence into repeatable command-line workflows.
ExecuteMove from an instruction toward an action in the terminal.
GovernKeep terminal AI usage within organisational controls.

From instruction to action

The CLI is the only surface that can change things on your machine, so the loop is deliberately explicit: it proposes, you confirm, then it acts.

01

In the terminal

Your instructionplain language, or a flag in a script
Repo & filesthe working tree you are standing in
Shell contextenvironment, branch, recent commands
Scripts & CIthe same binary, run unattended
becomes the task proposed change
02

The loop

Askinteract without opening a browser
Planstate the intended steps before taking them
Tool callruns under a sandbox profile you control
Diffshow exactly what would change
Confirmnothing lands on disk until you say so
every step audited result
03

The same governed path

Identity & accessterminal usage carries the same role and scope
Guardrailsprompts and output checked as everywhere else
Auditwhat ran, on whose behalf, and what it changed
prompt + repo context completion
04

Models

Model routerper task, not per product
Hosted by the providerAnthropic · OpenAI · Google, over the API
Open-weight, self-hostedfor air-gapped or restricted environments
GitHub repository Documentation
Licensing and governance

Open source, openly stated

AiNxt products are released as open source, sponsored by the National Payments Corporation of India. AiNxt Enterprise, AiNxt OS and AiNxt Code are published under the MIT License as single-vendor NPCI projects. AiNxt CLI is a fork of Grok Build, published under the Apache License 2.0 — its upstream copyright and provenance are separate from NPCI's. Contributions are accepted under a Developer Certificate of Origin rather than a contributor licence agreement — no paperwork, just a signed-off commit.

Provenance, stated plainly

AiNxt Enterprise, AiNxt OS and AiNxt Code are single-vendor projects: NPCI owns the copyright, employs the maintainers and makes the final call on scope, architecture and releases. AiNxt CLI is a fork of Grok Build — its upstream copyright belongs to the original authors; NPCI maintains the fork. Both are stated plainly because implying a provenance that does not exist would waste contributors' time.

Licences by product

AiNxt Enterprise, AiNxt OS and AiNxt Code are published under the MIT License — permissive, OSI-approved, no conditions beyond retaining the licence notice. AiNxt CLI is published under the Apache License 2.0 — also permissive and OSI-approved, with an explicit patent grant and a requirement to state changes made to the source.

Dependency posture

Every dependency installs from a hash-locked file with a pinned hash for each package, so a build either resolves to exactly the reviewed set or fails. Third-party licence notices are recorded in-tree, and tooling to generate a full CycloneDX software bill of materials ships with each repository.

Trademarks are separate

The licence covers code, not marks. Forks may state that they are built on AiNxt, but may not imply endorsement or official status from NPCI. If you fork it, name it something of your own.

Developer Certificate of Origin

Contributions are accepted under a DCO rather than a CLA — no legal paperwork, no individual or corporate agreement to sign. A Signed-off-by line in your commit is all that is required, certifying that you have the right to submit the contribution under the applicable open-source licence (MIT or Apache 2.0).

Security reporting

Each repository maintains its own security tracker. Vulnerabilities should be reported privately against the relevant repository rather than as public issues. The maintainer team triages and coordinates disclosure. Full detail is published alongside the source code in each repository's SECURITY.md.

Version one is earned, not scheduled. All four products are pre-1.0 and treat their public APIs as the contracts that matter. A product will not call itself stable until its interfaces have been implemented by someone outside the founding team. Until then there is no evidence the contracts generalise, and saying otherwise would waste implementers' time.
Developer ecosystem

Explore and build with AiNxt

AiNxt is intended to provide a path from using AI products to building new AI experiences. Each product has its own public repository on GitHub — browse the source, read the documentation and start building.

AiNxt OS

Foundation for building AI applications, agents, workflows and experiences.

GitHub repositoryDocumentation

AiNxt CLI

Command-line AI experience for developers and technical teams.

GitHub repositoryDocumentation

AiNxt Code

AI-assisted development directly inside the IDE.

GitHub repositoryDocumentation

AiNxt Enterprise

Enterprise AI experience across web and desktop.

GitHub repositoryDocumentation
Questions

Frequently asked questions

Are the four AiNxt products the same product?

No. AiNxt Suite is the umbrella for four distinct products: AiNxt Enterprise, AiNxt OS, AiNxt Code and AiNxt CLI. Each addresses a different way of working with AI.

What is AiNxt OS?

AiNxt OS is the foundational platform for developers and organisations that want to build their own AI applications, agents, workflows and experiences.

Is AiNxt OS the user interface for AiNxt Enterprise?

AiNxt OS and AiNxt Enterprise should be understood as distinct products. AiNxt OS provides a foundation for building AI experiences; AiNxt Enterprise provides an enterprise-facing experience across web and desktop.

Where can developers find AiNxt repositories?

Each product has its own public repository: ainxt-enterprise, ainxt-os, ainxt-code and ainxt-cli.

What is AiNxt Enterprise?

AiNxt Enterprise brings AI into enterprise work across web and desktop, including conversational, agentic, Office and software-development lifecycle experiences.

What licence does AiNxt use?

It depends on the product. AiNxt Enterprise, AiNxt OS and AiNxt Code are released under the MIT License. AiNxt CLI is released under the Apache License 2.0. Both are permissive, OSI-approved open-source licences sponsored by NPCI. Contributions are accepted under a Developer Certificate of Origin (DCO); no CLA or legal agreement is required beyond a Signed-off-by line in your commit.

Is AiNxt truly open source, or is it source-available?

AiNxt is genuinely open source. Enterprise, OS and Code are MIT-licensed single-vendor NPCI projects — NPCI owns the copyright and employs the maintainers. CLI is a fork of Grok Build, published under Apache 2.0 — its upstream copyright belongs to the original Grok Build authors; NPCI maintains the fork. None of the four are source-available, dual-licensed or behind a contributor agreement. What you see on GitHub is what it is.

Can I fork AiNxt and build my own product on it?

Yes. Both the MIT License (Enterprise, OS, Code) and the Apache License 2.0 (CLI) permit forking, modification and redistribution, including commercial use. Apache 2.0 additionally requires that you state changes made to the source and includes an explicit patent grant. In both cases you may not imply endorsement or official status from NPCI. If you fork it, name it something of your own and retain the applicable licence notice.