NexusGPT
NexusGPT is an enterprise platform for building and deploying autonomous AI agents without code, letting business teams connect internal systems, choose their own models, and publish agents to Slack, Teams, WhatsApp, phone or an API.
What is NexusGPT?
NexusGPT, marketed simply as Nexus and trading as nexusGPT, is an enterprise platform for building and running autonomous AI agents, aimed at business teams rather than developers. The address under review, gpt.nexus, is the application itself: a React single-page app sitting entirely behind Clerk authentication, with the public shop window kept on a separate marketing domain.
The vendor describes a three-layer arrangement. The bottom layer is the stack a company already runs, its models and business systems, with nothing to migrate. The middle layer is Nexus, where IT wires up systems, models and permissions once and turns them into reusable building blocks. The top layer is the agents themselves, built, deployed and owned by the team that uses them.
Inside the application sit agents, a chat playground, a shared inbox with user assignment and satisfaction scoring, workflows, triggers, documents and folders, Word and Excel document templates, custom tools with OAuth2 credentials, a code interpreter and model settings. Cue, announced in 2026, is an autonomous build loop: it scopes the use case with the user, asks clarifying questions, connects to internal data with the right permissions, builds the automation, tests it against realistic scenarios and publishes it to the team, in one to two hours.
Model handling is unusually open. Only recent commercial models with function calling work with agents, and the documentation names Mistral, Anthropic and OpenAI's GPT-4, but external models can be added and an open-source model can be hosted on the platform itself, with temperature and top_p exposed and a model attachable to a specific task.
Deployment is the widest surface. One agent can reach a website widget, the API, Slack, Microsoft Teams, WhatsApp, Telegram, Twilio, phone calls, Airtable and Facebook, as well as Excel, Word, PowerPoint, Outlook and Google Sheets add-ins. A public API is documented separately with webhooks and pagination, an admin console covers organisations, audit logs, rate limits, policies and observability, and the interface ships in six languages. The vendor advertises SOC 2, ISO 27001 and ISO 42001 certification and claims 174 agents in production for customers including Orange Group, Lambda.ai and Proximus Global.
What it does
- Build an autonomous AI agent without writing code
- Connect internal systems and third-party business tools to that agent
- Feed it a knowledge base from documents, Notion, SharePoint or a Google connector
- Choose, add or self-host the language model behind each agent
- Deploy the agent to a website widget, Slack, Teams, WhatsApp, Telegram, a phone line or the API
- Monitor conversations and take over from the agent in a shared inbox
- Govern access, permissions and audit logs across the organisation
When to use NexusGPT / When not to
A quick filter to help you decide if NexusGPT is the right fit.
When to use NexusGPT
- Large enterprises that want to industrialise AI agents without hiring a dedicated engineering team
- Non-technical business teams in sales, marketing, customer support, strategy and recruiting
- Regulated organisations that need audit trails, role-based access and documented governance
- IT departments that prefer to wire systems, models and permissions once, then let the business assemble agents on top
- Telecom and large service groups: Orange Group and Proximus Global are named customers, and the vendor runs a dedicated telco page
When not to use NexusGPT
- Individuals and small teams: there is no free plan, no free trial and no self-service sign-up
- Buyers who need a published price before talking to anyone, since pricing is per agent and quoted only after a demonstration
- Anyone wanting to evaluate the product first, as gpt.nexus sits entirely behind authentication
- Mobile-first users, because there is no native iOS or Android application
- Teams needing something live this week: the vendor itself quotes twelve weeks to an enterprise-scale agent in production
How to use NexusGPT
A typical end-to-end flow, from setup to results.
- Book a demonstration through the vendor's scheduling link, since there is no self-service sign-up
- Receive an account and sign in at gpt.nexus, which opens straight onto a new chat
- Have IT connect the systems, OAuth2 credentials, language models and user groups once, in the settings area
- Create an agent from the agents section
- Give it a knowledge base by uploading documents or connecting Notion, SharePoint or Google
- Attach the tools and custom actions the agent is allowed to call
- Test the agent in the chat playground against realistic scenarios
- Publish it to a channel, since an agent that is not deployed simply runs in the background
- Watch conversations in the shared inbox, assign colleagues, set statuses and read satisfaction scores
- Iterate with the Forward Deployed Consultants the vendor assigns for scoping and change management
Pros & Cons
Pros
- Deployment coverage is unusually wide: web, API, Slack, Teams, WhatsApp, Telegram, phone and Office add-ins from a single platform
- Three certifications on display, including ISO 42001, which is specific to AI management systems and still uncommon
- No lock-in to one model provider, since commercial, external and self-hosted open-source models are all supported
- Governance is built in rather than bolted on, with permissions, policies, audit logs and user groups
- A documented public API with webhooks and pagination, plus an embeddable widget
- Data retention is capped in writing at eighteen months after account termination
- Per-agent pricing does not inflate with headcount
Cons
- No price is published anywhere, so the tool cannot be compared without going through a sales call
- Neither a free plan nor a free trial is advertised, and there is no self-service sign-up
- Nothing is public on gpt.nexus itself, as the whole product sits behind authentication
- No Data Processing Agreement and no subprocessor list are published
- Data is hosted in the United States under Delaware law and exclusive Delaware jurisdiction, despite a largely European customer base
- The privacy policy has not been revised since 31 December 2023, and the product documentation still illustrates model choice with GPT-3.5 and GPT-4
- A single contact address, obfuscated on the page, is the only published written channel
Pricing & Plans
No free plan and no free trial are advertised, and no price is published. The vendor states that pricing is charged per agent and depends on the processes being automated, so no entry-level amount or currency can be quoted. The terms describe a subscription that may be cancelled at any time, taking effect at the end of the current paid term, with all purchases non-refundable. A figure is obtained only after requesting a demonstration.
Data, GDPR & hosting
A consolidated view of how NexusGPT handles your data.
GDPR overview
The privacy policy carries a dedicated section for users in the EU and the United Kingdom, stating that the GDPR and UK GDPR require it to explain its legal bases, which it then lists: consent, performance of a contract, legitimate interests, legal obligations and vital interests. People in the EEA, the United Kingdom, Switzerland and Canada are granted access, rectification, erasure, restriction, portability, objection and freedom from automated decision-making, with complaints directed to a member-state authority, the UK regulator or the Swiss Federal Data Protection and Information Commissioner. A Data Protection Officer is named under Article 37. However, no Article 27 representative is designated, no Data Processing Agreement is published, and the phrase "GDPR compliant" never appears: the rating reflects the mechanism on display rather than a claim by the vendor.
Who owns the data?
The terms and the privacy policy are published by NEXUS ENTERPRISES and incorporate one another. Material you submit falls under the terms' User Generated Contributions and Contribution License clauses, and a separate User Data section governs what you upload. The only third-party sharing situation the privacy policy lists is a business transfer, namely a merger, a sale of company assets, financing or an acquisition. The vendor states that it does not retain data obtained through your connected Workspace, such as Google APIs, to develop, improve or train generalised AI or ML models. Access, correction and deletion requests go through an online form, and a named Data Protection Officer handles queries about the notice.
Reuse rights
The privacy policy sets out why the vendor processes information: delivering and administering the service, managing accounts, sending marketing messages you can unsubscribe from at any time, fraud monitoring and security, analysing usage trends to improve the product, and protecting vital interests. Registering through a social account passes profile details from that provider, often a name, an email address, a friends list and a profile picture. Cookies, web beacons and pixels are used, with a separate Cookie Notice, and approximate location is inferred from your IP address unless you refuse it. On the customer side, the vendor commits to not training generalised AI or ML models on data obtained through your connected Workspace, and nothing published sets out a broader reuse licence over your business content.
Data retention & training
Hosting summary
The terms state plainly that the services are hosted in the United States, and warn users elsewhere that accessing them means transferring their data there voluntarily. No alternative hosting region is offered and no data-residency choice is documented publicly, although the vendor's sales material mentions data-residency controls as a capability. Legal disputes fall under Delaware law and the exclusive jurisdiction of Delaware courts. Technically, the domain runs on Cloudflare DNS with the front end served by Vercel, the apex address geolocates to the United States on Amazon infrastructure, and error reporting goes to a European Sentry region. Those are indications about the delivery chain rather than a statement about where customer data comes to rest. No subprocessor list is published, so the third parties involved in processing cannot be enumerated from public sources. Buyers with European data-residency obligations should treat the United States as the answer and obtain anything else in writing.
Things to keep in mind
Risks and trade-offs to weigh before adopting NexusGPT.
- Pricing opacity makes budget control difficult: the cost only emerges after a sales conversation, and all purchases are non-refundable
- Data is hosted in the United States and disputes fall under exclusive Delaware jurisdiction, which European buyers should weigh before signing
- No Data Processing Agreement and no subprocessor list are published, so both have to be demanded contractually rather than read beforehand
- The commitment not to train models is worded only for data obtained through your connected Workspace, not for all customer data
- Agents acting autonomously on systems of record can propagate a mistake at scale before anyone reviews it; the audit trail records what happened, it does not prevent it
- Handing routine judgement to agents can quietly erode a team's own grasp of the process it automated, leaving nobody able to check the output
- The Delaware address is a registered-agent address while the operating team is in Brussels and San Francisco, which matters for escalation and accountability
Setup & Integrations
Technical difficulty
Low for the business user, moderate for IT. No development skill is needed to assemble an agent, and Cue can scope, build, test and publish an automation in one to two hours. IT does the real work once, connecting systems, OAuth2 credentials, language models and user groups. The friction is organisational rather than technical: there is no self-service sign-up, so nothing starts before a demonstration, and the vendor itself quotes twelve weeks to an enterprise-scale agent in production, with its own consultants handling scoping and change management.
Deployment
Integrations
Supported languages
Behind NexusGPT
Fundraising
Social
Resources
All the official URLs gathered for verification and reference.
Alternatives
Tools that compete with or complement NexusGPT.
Frequently asked questions
What is NexusGPT?
Do I need to know how to code?
How much does NexusGPT cost?
Is there a free trial or a free plan?
Where can an agent be deployed?
Which language models can it use?
Where is the data hosted, and is it used for training?
How long is data kept?
Which certifications does the vendor hold?
Who is behind NexusGPT?
Should you pick NexusGPT?
NexusGPT is squarely an enterprise product, and it does not pretend otherwise. Everything about it, from per-agent pricing quoted only after a call to the absence of any free tier and the fact that gpt.nexus shows a sign-in form and nothing else, signals a tool bought by a committee rather than tried on a whim.
What it buys is breadth and control. The deployment surface is genuinely wide: one agent can be published to a website widget, an API, Slack, Microsoft Teams, WhatsApp, Telegram, a phone line, or an Excel or Outlook add-in. Model choice is open, up to hosting an open-source model on the platform. Governance is treated as the foundation, with user groups, access tiers, policies and audit logs, and the vendor backs that with SOC 2, ISO 27001 and ISO 42001 certification, the last of which covers AI management systems and is still uncommon.
The gaps are just as clear. No Data Processing Agreement and no subprocessor list are published, a strange omission for a product sold to regulated European enterprises whose data is hosted in the United States under exclusive Delaware jurisdiction, and no Article 27 representative is designated. The privacy policy has not been touched since December 2023, and the documentation still explains model choice through GPT-3.5 and GPT-4, which says more about upkeep than about the platform but which a buyer will notice.
The company is young, founded in 2024 and funded with a seed round of USD 4.3 million, yet it already names Orange Group, Lambda.ai and Proximus Global as customers and publishes precise outcome figures. Those figures are the vendor's own and unaudited. Treat NexusGPT as a serious contender for a governed, multi-channel agent programme, and put the missing contractual paperwork on the agenda of the first call.
- Choosing a selection results in a full page refresh.
- Opens in a new window.