GPTBots
GPTBots is an enterprise platform for building, deploying and running AI agents without code. It covers customer support, sales development, enterprise search and data analysis, connects to more than a dozen model providers, and offers private on-premise deployment.
What is GPTBots?
GPTBots is an enterprise platform for building, operating and scaling AI agents without writing code. It is published by Metaverse Cloud Pte. Ltd. together with GPTBots Hong Kong Limited, which present themselves as part of the Aurora Mobile group, listed on NASDAQ under the ticker JG.
The product is organised around two spaces. DevSpace is where builders assemble agents, workflows, tools and models. WorkSpace is where employees consume them, through search, published agents, workflows and an internal AI marketplace. Four agent types are documented: Agent for straightforward scenarios, FlowAgent for hand-designed logic, LoopAgent, and MultiAgent with handoff to a human operator.
Knowledge is the backbone. The knowledge base ingests doc, docx, pdf, txt, markdown, csv and spreadsheet files, crawled web pages, question-and-answer pairs and Google Drive documents, then applies parsing strategies per file type, hybrid sparse-and-dense vector retrieval, query augmentation and reranking, with editing down to the individual chunk. Structured data is covered too: agents connect to MySQL, PostgreSQL, SQLite, SQL Server, Oracle, MongoDB and Redis, turn natural-language questions into joined SQL queries, and render the results as interactive charts.
The platform is deliberately model-agnostic. A published Model Services table names OpenAI, Anthropic, Google, Meta, Azure, Alibaba, Baidu, Tencent, Mistral and DeepSeek among others, and customers may either use the vendor's keys or bring their own, which lowers the credit charge. Fine-tuning corpora can be generated from knowledge-base content and past conversations.
Four packaged solutions are marketed: customer support, AI SDR, enterprise search and data insights. The support page claims a 90% AI resolution rate, a 21% human handover rate and coverage of more than 90 conversational languages. Agents publish to a bubble widget, an iframe, WhatsApp, Slack, Discord, Telegram, Messenger, Instagram, LINE, Microsoft Teams, Intercom, LiveChat and Zapier, or are called through a REST API covering conversations, knowledge, metadata, workflows and databases. Audio agents add speech recognition, speech synthesis and telephone integration. Regulated buyers can move to a private cloud-native or on-premise deployment carrying their own branding and domain.
What it does
- Build an AI agent by drag-and-drop, with no coding required
- Feed it a knowledge base from documents, spreadsheets, crawled pages, Q&A pairs or Google Drive
- Orchestrate multi-step workflows that chain several language models and tools
- Publish the agent to a website widget, WhatsApp, Slack, Discord, Telegram or Microsoft Teams
- Query enterprise databases in plain language and turn the answers into interactive charts
- Hand a conversation over to a human operator when the agent reaches its limits
- Drive the whole platform through a REST API or embed it with the iOS and Android SDKs
When to use GPTBots / When not to
A quick filter to help you decide if GPTBots is the right fit.
When to use GPTBots
- Customer service organisations automating high-volume enquiries, where the vendor claims a 90% AI resolution rate and a 21% human handover rate
- Sales development and lead generation teams capturing prospects through WhatsApp, Facebook, Telegram and other messaging channels
- Companies sitting on scattered internal documentation that they want to make searchable and answerable by every role
- Data teams that need business users to query MySQL, PostgreSQL, Oracle or MongoDB in plain language and get charts back
- Large enterprises, financial institutions and public-sector bodies that require a private cloud-native or on-premise deployment
When not to use GPTBots
- Individual consumers: the platform is contractually restricted to people aged 18 or over and to registered legal entities
- Buyers who need a published subscription price before they talk to anyone, since only the free tier and the credit top-up rate are public
- Organisations that require European data residency by default, as data sits on servers in Singapore with no Article 27 representative appointed
- Teams looking for a ready-made general-purpose assistant rather than a platform on which agents must be designed and trained
- Any use case that depends on guaranteed accuracy, or on professional medical, legal or financial advice, both of which the service agreement explicitly excludes
How to use GPTBots
A typical end-to-end flow, from setup to results.
- Create a free account from the Join Free link; the free tier grants 100 credits every month
- Create an organisation and pick its data centre at that moment, Singapore being the default
- Start from a sector template or from a blank agent
- Import your business knowledge: files, a crawled website, question-and-answer pairs or Google Drive documents
- Configure persona, tone, constraints, memory, collection forms, tools and workflows
- Test the agent in a live conversation and correct its answers directly in training mode
- Connect the channels you need, from the bubble widget and iframe to WhatsApp, Slack, Teams or Intercom
- Publish the agent and monitor conversations, question categories and sentiment in the logs
- Track credit consumption under Organization - Usage, and download receipts under Organization - Billing
- For deeper integration, call the REST API or embed the agent with the iOS and Android SDKs
Pros & Cons
Pros
- Model-agnostic by design: more than a dozen LLM providers are integrated, and customers can bring their own keys
- Unusually broad scope for one platform, spanning agents, workflows, retrieval, databases, audio and analytics
- Credit pricing is published model by model, down to the cost per thousand tokens
- The free tier is permanent rather than a countdown trial, with 100 credits granted every month
- Twenty-seven documented integrations, including WhatsApp, Slack, Microsoft Teams, Intercom and Zapier
- An explicit contractual commitment not to train on customer data without separate opt-in consent
- Private on-premise deployment is available for regulated sectors, alongside ISO 27001, ISO 27701 and SOC 2 Type II certifications
Cons
- No public pricing page at all: the cost of the Business and Enterprise plans is only available by emailing the vendor
- The only published rates live in the developer documentation rather than on the marketing site
- Credits are non-refundable, non-transferable and expire: monthly credits after thirty days, topped-up credits after one year
- Paid subscriptions cannot be cancelled, refunded or transferred once activated, and late payment carries a 0.05% daily penalty
- Data is hosted in Singapore by default, with no customer-facing DPA and no EU Article 27 representative
- Headline performance figures of 70%, 90% and 300% are published without methodology or verifiable source
- There is no mobile app on either store; mobile use goes through an integration SDK that a developer must implement
Pricing & Plans
A permanent free plan is available to all users and grants 100 credits per month. The lowest published paid price point is a credit top-up of USD 10.00 for 1,000 credits, valid for one year and payable once through Stripe, on the stated basis that 100 credits equal one US dollar. The price of the Business and Enterprise subscription plans is not published; prospective subscribers are asked to contact the vendor by email.
- open to all users
- 100 credits per month
- intended for evaluation and testing
- 10
- 000 credits
- up to 100 agents and workflows created
- up to 10 published
- 10 team seats
- advanced features and a higher rate limit
- price not published
- agent development
- agent operations
- private deployment options
- AI project consulting and implementation services
- price not published
- available on every plan
- 1
- 000 credits for USD 10.00
- valid one year
- one-off payment
Data, GDPR & hosting
A consolidated view of how GPTBots handles your data.
GDPR overview
The privacy policy, effective 20 May 2025, devotes a closing section to the European Economic Area, but it is written mainly to push obligations onto the customer: a partner processing personal data must itself comply with the GDPR if it is established in the EEA, offers goods or services there, or monitors behaviour there. For EEA individuals the vendor does set out its own legal bases, naming contract performance, legitimate interest, consent, legal obligation and vital interests. Data subject rights are described in detail, with a fifteen working day response commitment. The vendor claims ISO/IEC 27001, ISO/IEC 27701, SOC 2 Type II and ESOF certifications. There is, however, no explicit statement that GPTBots itself is GDPR-compliant, no Article 27 representative named, and no data protection officer identified anywhere on the site.
Who owns the data?
The service agreement is explicit on this point. Section 3.1.1 states that both the Input Content submitted and the Output Content generated belong to the customer or to the rightful owner, and that using the service does not change that ownership. Section 3.5.1 extends the same rule to uploaded files and data, and grants the customer control over access, modification and deletion within the functional limits of the product. Data is treated as confidential information and is not disclosed to third parties, except to the third-party model providers needed to run the service. Ownership of the platform itself, by contrast, remains entirely with the vendor and its affiliates.
Reuse rights
Customers may reuse their own input and output freely, since the agreement leaves ownership with them and asks only that applicable law be respected and that AI-generated output be labelled as such when shown to third parties. The vendor's own permitted uses are narrower and enumerated. Section 3.5.2 limits processing to providing, maintaining and improving the subscribed service, and to passing content on to the third-party large language model providers that execute the request. Training is ring-fenced: customer data is not used to train the vendor's general public models or any other customer's models unless the customer gives explicit, separate, opt-in consent. Selling, renting or leasing data is prohibited, internal employee access is limited to staff who need it, and Google Workspace content is excluded from generalised model training. One nuance deserves attention: inside WorkSpace, the creator of a bot may view all conversation data of the people using it.
Data retention & training
Hosting summary
Customer data and end-user data are stored, in principle, on servers located in Singapore, and the storage jurisdiction stated in the service agreement is Singapore law. The data centre is selected when the organisation is created on the platform, Singapore being the default, and the documentation presents that choice as a way to meet compliance requirements. The privacy policy allows transfers outside Singapore for business needs, requiring equivalent protection from the recipient, and registered users are deemed to consent to cross-border transfer and to bear its risks. A significant carve-out applies to prompts and uploads: content routed to third-party language model providers is transmitted to and stored in their own data centres outside Singapore, governed by their published rules. The domain resolves to an address in Singapore hosted on Huawei Cloud. Announced safeguards include SSL in transit, encryption at rest, access control, data isolation and disaster recovery. Private deployments can instead run on a private cloud such as AWS or Microsoft Azure, or on the customer's own servers.
Things to keep in mind
Risks and trade-offs to weigh before adopting GPTBots.
- Credits are a one-way commitment: they are non-refundable, non-transferable and non-convertible, monthly credits vanish after thirty days and topped-up credits after one year
- A paid subscription cannot be closed, refunded or transferred once activated, and unpaid balances accrue a 0.05% penalty per day of delay
- Everything typed into an agent is forwarded to third-party model providers and stored in their data centres outside Singapore, under their rules rather than the vendor's
- Inside WorkSpace, the creator of a bot can read every conversation its users hold with it; declining that visibility means losing the feature
- Teams that embed the platform in their own product carry the responsibility for their end users' consent, including consent to the transfer of data to Singapore
- The vendor disclaims any guarantee that generated output is accurate, and contractually requires customers to label it as AI-generated and to review it before automated decisions
- An account left unused for twenty-four consecutive months may be closed by the vendor, and an export control clause bars use towards embargoed countries that the agreement never names
Setup & Integrations
Technical difficulty
Low for the mainstream path. The builder is drag-and-drop, prebuilt templates deploy in one click, and the vendor claims an agent can be assembled in about five minutes through a four-step route: create, debug, deploy, then train. Channel connections are documented step by step. Difficulty rises for the developer track, where the REST API, the mobile SDKs, custom tools and database connectors all require engineering skills. A private deployment is a project in its own right, involving hardware and software preparation with the vendor's team, announced as roughly two weeks.
Deployment
Integrations
Supported languages
Behind GPTBots
Social
Resources
All the official URLs gathered for verification and reference.
Alternatives
Tools that compete with or complement GPTBots.
Frequently asked questions
Is there a free plan?
How much does the Business plan cost?
How do credits work?
Is my data used to train AI models?
Where is the data hosted?
Which language models can I use?
Is there an API?
Where can an agent be published?
Is there a minimum age?
Can the platform be deployed on our own infrastructure?
Should you pick GPTBots?
GPTBots is one of the more complete enterprise agent platforms available today, and its scope is its main argument. A single product covers no-code agent building, workflow orchestration, retrieval over unstructured documents, natural-language querying of production databases, voice agents and operational analytics, and it refuses to lock buyers into one model vendor: more than a dozen providers are integrated, and customers who already hold their own API keys pay a reduced service charge for using them. The contractual language on data is clearer than most, with ownership left to the customer and training ring-fenced behind an explicit opt-in.
The reservations are commercial rather than technical. There is no pricing page: the only public figures are the permanent free tier of 100 credits per month and a top-up rate of USD 10.00 for 1,000 credits, and everything above that requires an email exchange. Credits expire, subscriptions cannot be cancelled or refunded once activated, and the headline efficiency claims scattered across the site come without any published methodology. Buyers subject to European rules should also weigh the default hosting in Singapore, the absence of an Article 27 representative and the lack of a customer-facing data processing agreement.
For an enterprise willing to run a proof of concept on the free tier and negotiate terms directly, GPTBots is worth a serious look, particularly where a private or on-premise deployment is a hard requirement. For anyone who needs a transparent, self-service price before committing engineering time, the evaluation will stall early.
- Choosing a selection results in a full page refresh.
- Opens in a new window.