Graphext
Graphext is a no-code visual analytics platform turning messy datasets into interactive charts, network graphs and predictive models. It connects to data warehouses and uses AI to surface topics, sentiment and entities for business teams, not data scientists.
What is Graphext?
Graphext is a visual advanced-analytics platform built for people who need answers from data without writing code. Its stated ambition is that data insights should not require a PhD: users upload or connect a dataset and immediately start exploring patterns, clusters and relationships through a fast, highly interactive interface.
The interface adapts to the shape of the data. Graphext detects column types on import and unlocks different operations accordingly, covering basic types such as numeric, categorical, ordinal, URL and date values, and advanced types such as free text, lists and graph links. Text columns unlock natural language processing, embeddings and sentiment analysis; numeric columns unlock statistical tooling.
Exploration rests on linked cross filters, a Compare tab, a Correlations tab, advanced filter queries and saved segments. Visualisation is handled by engines the company wrote itself: a WebGL and GPU-accelerated graph renderer designed for millions of nodes, and an extended version of Vega for charting with automatic chart recommendation and publication-quality export.
Generative AI runs through the workflow rather than sitting beside it. Graphext can auto-tag variables, create new columns, discover topics automatically, extract entities, classify content, detect intent and sarcasm, and answer questions through an AI chat. On top of that sits predictive modelling: classification, regression, clustering and time series, built by selecting a target variable, with feature-importance charts explaining each prediction.
Technically the platform is hybrid. Light steps run in the browser through WebAssembly with the in-house DARRO engine written in C++ and JavaScript; heavy steps run in cloud Docker containers through the BEIRO Python engine, with GPUs, autoscaling and Apache Arrow transport. Every action taken in the interface is serialised into a constrained domain-specific language called Recipe DSL, so analyses are reproducible, versionable and shareable as reusable recipes and project templates.
Sixteen data connectors are documented, from BigQuery, Snowflake and Databricks to S3, Google Sheets and Notion, with write-back where permissions allow. Published use cases span market and social media research, surveys, customer support analysis, advertising, data quality and defence or cyber intelligence work. The publisher is a Madrid-based branch of Graphext Inc, founded by Victoriano Izquierdo and Miguel Canton.
What it does
- Explore any dataset visually through linked cross filters, where selecting in one chart updates every other view instantly
- Enrich free text with AI to extract topics, sub-topics, sentiment, entities, intent and tone of voice at scale
- Build network graphs rendered on the GPU to connect entities and reveal clusters and emerging patterns
- Train predictive models for classification, regression, clustering and time series in a few clicks, with explainable feature importance
- Connect and refresh data from warehouses, databases, object storage and spreadsheets, on a schedule down to hourly
- Clean, transform and auto-tag columns using generative AI, and ask questions of the data through an AI chat
- Export publication-ready charts and datasets, or write results back to a connected integration
When to use Graphext / When not to
A quick filter to help you decide if Graphext is the right fit.
When to use Graphext
- Data and business analysts who need to explore large datasets visually without writing SQL or Python
- Market researchers and insights teams working on social media, news and survey data
- Customer experience and support leads segmenting tickets and conversations to find escalation drivers
- Growth, performance and AdTech marketers building contextual segments and explaining campaign performance
- Intelligence, defence and cyber analysts who need to connect entities and spot emerging patterns in text
When not to use Graphext
- Engineering teams that need to call the tool programmatically: no public API, documentation or endpoint could be found
- Buyers who must budget before talking to sales, since no price is published anywhere on the site or in the documentation
- Organisations requiring a signed DPA or a published subprocessor list, neither of which the vendor publishes
- Teams needing data residency outside Belgium, as Google Cloud Europe is the only hosting location documented
- Anyone looking for a mobile experience: no iOS or Android application exists and the platform runs in the browser only
How to use Graphext
A typical end-to-end flow, from setup to results.
- Create a free account at accounts.graphext.com, which the licence describes as free of charge to open
- Upload a data file, or connect one of the sixteen documented sources such as BigQuery, Snowflake, PostgreSQL, S3 or Google Sheets
- Optionally schedule a recurring refresh so the project picks up new data, at a frequency as short as one hour
- Let Graphext detect column types on import, which determines the operations, charts and models offered
- Explore visually with cross filters, the Compare and Correlations tabs, advanced filter queries and saved segments
- Enrich text columns with AI to extract topics, sentiment, entities and intent, supplying your own third-party API keys where required
- Train a predictive model by picking a target variable, then read the feature-importance chart to understand the drivers
- Start from a project template such as churn analysis, lead scoring or customer segmentation to save setup time
- Save findings as Insights and build publication-ready charts with the chart composer
- Export the results as CSV or Excel, or write them back to an integration configured with write access
Pros & Cons
Pros
- Genuinely no-code: exploration, enrichment and modelling all run without SQL or scripting
- A permanent free plan with real capacity, allowing four projects of up to 4,000 rows each and three invited team members
- AI features are included in the pricing rather than sold as an add-on, according to the vendor's documentation
- Purpose-built rendering engines, using WebGL and GPU acceleration in the browser and autoscaling cloud compute for heavy steps
- Full reproducibility, since every interface action is serialised into code that can be versioned and reused
- Explicit European hosting on Google Cloud in Belgium, with encryption at rest and audit logging
- Substantial public documentation, with hundreds of indexed pages and detailed FAQs on security, performance and data sources
Cons
- No price is published anywhere, and the pricing page the documentation links to twice returns a genuine 404
- No public API could be found, despite the product page advertising that models can be exported via API
- The OpenAPI specification announced in the documentation index returns a 404 identical to a control request
- The SOC 2 and ISO badges in the footer carry no certificate number or certifying body and are absent from the site's own compliance page
- Neither a Data Processing Agreement nor a subprocessor list is published
- No documented way to opt out of AI training, and no stated position on whether customer data is used for training
- Several site links are broken, including the footer blog link, and no mobile application exists
Pricing & Plans
A permanent free plan is available: account creation is free of charge and the user licence caps it at 4,000 rows processed per project, four projects per account and three invited members per shared team. Beyond that, the vendor's documentation describes a Pro plan with almost no limitations and an Enterprise plan negotiated with the sales team, which may include professional services. No price is published for either paid plan. Amounts are set on an Order Form, are exclusive of tax with VAT added, and are non-refundable; the contract runs for twelve months and renews annually unless notice is given ten days before the end of the period. The pricing page referenced twice by the documentation does not exist, so no entry price can be quoted.
- free account creation
- 4
- 000 rows processed per project
- 4 projects per account
- 3 invited members per shared team
- price: free
- described by the vendor as offering additional features and almost no limitations
- no price published
- adapted to specific needs through the sales team
- includes professional services
- dedicated support and onboarding assistance
- no price published
Data, GDPR & hosting
A consolidated view of how Graphext handles your data.
GDPR overview
GDPR implementation is documented rather than merely claimed. The privacy policy, last modified on 13 June 2024, names Graphext Inc Sucursal en Espana as controller, gives its Madrid address and NIF W0279624A, and designates a Data Protection Officer reachable at dpo@graphext.com. It lists the six data subject rights, states that transfers outside the EEA rely on Standard Contractual Clauses or adequacy decisions, and points users to the Spanish data protection authority, the AEPD, for complaints. The documentation states the company successfully went through an audit process to be compliant with GDPR and links an attestation. Services are not intended for anyone under 14. No Article 27 representative is named, which is consistent with a controller established in Spain. No Data Processing Agreement and no subprocessor list are published.
Who owns the data?
The user uploads the data to be analysed, and the vendor's documentation states that copies are stored encrypted at rest in Google Cloud Storage, with the user keeping absolute control over those copies and able to remove them at any time. Vendor employees access customer data only after the user's explicit consent; access is limited to a small group of trained staff, allocated by role and logged for every request. Ownership of the software itself stays entirely with the publisher: the user licence grants a non-exclusive, non-transferable right of use and reserves all intellectual and industrial property to Graphext. The declared data controller is Graphext Inc Sucursal en Espana, NIF W0279624A, in Madrid.
Reuse rights
The licence grants an end-user right to access and use the software, not a right over the vendor's technology: the user may not transfer, sublicense, modify, reverse engineer, decompile or disassemble it. Data the user brings in, however, remains the user's to work with and to take away. The documentation describes two export routes, downloading the original dataset or a filtered selection as CSV or Excel, and writing results back to any connected integration for which write access has been enabled, so outputs such as predictive scores can be pushed to a warehouse. Personal data collected through the website is processed for contractual purposes, legitimate interest, consent and legal compliance, and may be shared with public authorities, service providers and business collaborators. Nothing in the site or the documentation states whether customer data is used to train AI models: the subject is never addressed.
Data retention & training
Hosting summary
The platform is hosted on Google Cloud in Europe, in Belgium, at the Saint-Ghislain data centre, according to the vendor's documentation. That statement is independently consistent with what the application hosts resolve to, namely Google infrastructure in Brussels. Datasets are stored in a private Google Cloud Storage bucket, encrypted at rest, and served to authenticated users through signed URLs valid only for a very short period. The cluster is described as disconnected from the internet except for port 443, with all web traffic over SSL and port 80 redirected; private key authentication is required to administer it. User analysis jobs run on isolated machines on a separate network with no direct access to internal systems, and Google Cloud Audit Logs monitor the infrastructure. The marketing website itself is served from a separate content delivery network and should not be confused with where customer data lives. No alternative region or data residency option is documented, and the privacy policy notes that transfers outside the EEA rely on Standard Contractual Clauses or adequacy decisions.
Things to keep in mind
Risks and trade-offs to weigh before adopting Graphext.
- Advertised capability that could not be verified: the product page promises model export via API, but no API, endpoint or specification exists, so do not build an integration plan on it
- Unsupported trust signals: the SOC 2 and ISO footer badges carry no number or certifying body and are absent from the site's own compliance page, which documents only a GDPR audit
- Total price opacity invites commitment before cost is known, on a twelve-month contract that renews automatically unless notice is given ten days ahead
- Interface figures such as topic and sentiment percentages, and the social profiles shown in the market research walkthrough, are mock-ups; treating them as benchmarks would set false expectations
- Automated topic and sentiment labelling can create false confidence: the tool infers meaning statistically, and a user who stops sampling the underlying text may act on a category that does not hold
- No published Data Processing Agreement, subprocessor list or training opt-out, which matters if you upload personal data or customer conversations
- The publisher's legal documents predate the contracting entity, with a legal notice from 2020 and a licence from 2022 against a company registered in November 2023, and three different postal addresses appear across the site and the register
Setup & Integrations
Technical difficulty
Low for a first project, moderate to industrialise. A free account is opened self-service and nothing is installed: the platform runs entirely in the browser. Uploading a file and exploring it requires no technical skill, and project templates for churn, lead scoring and segmentation shorten the first run. Effort appears at the edges: warehouse connectors need credentials, and explicit write permissions if you want results pushed back, while AI enrichment requires you to supply your own third-party API keys. The Recipe DSL is optional, since the interface serialises the code by itself.
Deployment
Integrations
Supported languages
Behind Graphext
Fundraising
Social
Resources
All the official URLs gathered for verification and reference.
Frequently asked questions
What kinds of text data can Graphext analyse?
Do I need coding or data science skills to use it?
Is there a free plan, and what are its limits?
How much does the paid version cost?
Does Graphext offer an API?
Which languages does it support?
Which tools and data sources can I connect?
Where is my data hosted and how is it protected?
Can Graphext employees see my data?
Is there a minimum age to use the service?
Should you pick Graphext?
Graphext is a technically serious product with an unusually candid free tier and a genuinely no-code promise that the feature set supports. The in-house WebGL graph renderer, the WebAssembly and cloud hybrid architecture and the Recipe DSL that serialises every click into reproducible code are not marketing veneer: they are documented in detail, and the public documentation runs to hundreds of indexed pages. For an analyst or a market researcher who needs to make sense of large text datasets without a data science team, the offer is coherent and the free plan is substantial enough to evaluate it properly.
The reservations are commercial and evidential rather than technical. No price is published anywhere, and the pricing page the vendor's own documentation links to twice does not exist, so any budgeting requires a sales conversation. The product page advertises exporting models via API, yet no API, documentation or specification could be found, and the OpenAPI file announced in the documentation index returns an error. The SOC 2 and ISO badges in the footer carry no certificate number or certifying body and are not repeated on the site's own compliance page, which documents only a GDPR audit. No Data Processing Agreement or subprocessor list is published, and the question of training on customer data is never addressed.
There is also a chronology worth knowing: the contracting entity, a Spanish branch of Graphext Inc, was entered in the Madrid commercial register on 6 November 2023, while the site dates the product to 2019 and the domain to 2014. None of this makes the tool less capable, but a buyer with procurement or compliance obligations should expect to ask for in writing what the site does not publish.
- Choosing a selection results in a full page refresh.
- Opens in a new window.