BI.P.EYE logo
Signup

The data layer for AI & embedded analytics

Accurate answers from all your company's data, for every team. In Claude, Copilot, your product or your dashboards. AI sets it up, your team approves.

BI.P.EYE connects every source into one data layer, proposes the model, metrics and access rules with AI, and tests itself on real questions. Your team reviews and approves each change, and every number in an answer is traceable.

✦ AI Semantic Layer ✦ Model Training Agents Multi-Source Data API-First Every Number Traceable Claude MCP
Get started How to start
Security certified: ESOF AppSec ADA CASA
An independent application security assessment by TAC Security, benchmarked against the OWASP Application Security Verification Standard, and re-assessed annually.
Contact us for any Data/BI project - [email protected]
ASK YOUR DATAFrom question to trusted answer

Trusted by product and data teams

Amsalem
Swap
Everywatch
EJ Capital
MutualArt
IBI
In production
Live since 2023

At companies across different industries

Monthly users
3K+ live users

Every month, on dashboards, embeds and chat

Three ways customers use it
Embedded analytics Data layer AI readiness

Inside their own products, as the base for all their data, or to prepare it for AI

Auto models · AI semantic layer

An AI semantic layer that builds and trains itself. Your team approves.

Connect your sources and AI prepares the data, builds the semantic layer, trains it with suggested questions and keeps sending AI suggestions to improve it. Your team reviews and approves every change before it goes live.

  • AI semantic layer
  • AI data prep
  • Auto data modeling
  • Semantic layer training
  • AI suggestions
  • Human-in-the-loop
  • Governed metrics
  • Row-level security
  • Version control
1

Data prep

Connects your sources and suggests merges, joins, cleaning and where to store

  • No code
  • You approve
2

AI semantic layer

Generates metrics, dimensions, joins and access rules from your data

  • Checked against real values
  • Hundreds of tables
3

Suggested questions

Questions your data can answer, for a model, a dashboard or a chart

  • One click asks it
  • Field names, never rows
4

Training the model

Each test shows how the answer was built; your team marks it

  • Admin approves
  • Every version restorable
No code to connect and prepare No manual mapping to start Nothing changes without approval Every version can be restored
Accuracy

Correct answers that get better with every question

Answers use your own definitions, not guesses. Each one shows how it was reached and says what it could not answer, and the model keeps training itself on your team's questions, so answers get more accurate over time.

How many orders and net revenue did we have by month and country?
Month Country Orders Net revenue
2026-09 United Kingdom 43,770 $2.44M
2026-08 United Kingdom 41,208 $2.31M
2026-08 United States 28,915 $1.87M
How we got this answer
SOURCEOrders
PERIODLast 12 months
BYMonth and country
COUNTEach order once
VALUENet revenue, your definition
  • One row per month and country.
  • Each order is counted once, even with several product lines.
  • Net revenue uses your approved definition, the same as your dashboards.
  • You also asked about refunds by reason. That is not in this data, so the answer says so.
Illustrative example

Uses the values you really have

Filters match what is actually in your data, so "active customers" means the customers you count as active.

Totals you can trust

Nothing is counted twice, even when the data comes from several places.

Honest about what it can't answer

If part of a question is not in your data, the answer says so instead of guessing.

Trains itself to answer better

The model learns from the questions your team asks and suggests improvements. Your team approves, and answers get more accurate over time.

Claude MCP server

Ask questions about your company's data directly in Claude

The BI.P.EYE MCP server connects Claude to your models, relationships and semantic layer, so answers use the same definitions as your dashboards.

  • Claude queries your BI.P.EYE models - it never connects to your databases directly
  • Large results, like a 200k-row table, come back as a download link for Excel
  • Performance, caching, security and query efficiency are handled on the BI.P.EYE side
Add to Claude Setup guide

Listed in the Claude connectors directory. Sign in with your BI.P.EYE account.

Claude
+ BI.P.EYE MCP server
Why did our revenue drop last month?
bipeye · query_model("Commerce")

Net revenue fell 12% vs. the prior month. Most of the gap comes from two channels:

Channel Prior Last month Change
Paid social $184k $139k -24%
Email $96k $81k -16%
Organic $212k $214k +1%
Illustrative example
AI readiness

Prepare your data for AI

An AI answers from what it knows about your data. Column names alone rarely say what a field means, which table to join, or how revenue is calculated. BI.P.EYE scores every dataset and model on how much of that is written down, shows what is missing, and helps you fill it in.

1

See how ready each dataset and model is

A readiness score on every dataset and model, built from field and dataset descriptions, the model description, relationship meanings and defined metrics. The next step is always one click away.

2

Fill the gaps with AI, then review

Generate descriptions for datasets and fields, field roles, metrics, a glossary and example questions. Drafts only fill what is empty, and nothing is saved until you accept it.

3

Keep restricted data out of it

Values your access rules restrict and personal data are masked before the AI sees a sample, and cannot be saved into the layer. Each user, API key and embedded customer only gets the parts about datasets they can access.

The same layer is used by the AI chat on dashboards, embedded chat in your product, and Claude.

Check your data's readiness How models work
How it works

One platform, from raw data to trusted answers

Connect any data, model it once, and use the same definitions everywhere: in dashboards, inside your product, in chat and in Claude.

Data layer

Different sources, one unified data layer

No-code pipelines pull from each source, reshape the data, match records across systems, and store the result in the destination you choose. Authentication, pagination, rate limits, retries and nested JSON are handled for you.

AI Assistant for Your Data Live 

Use the AI assistant to analyze your data. Embed it in your website or app so your customers can analyze their own data with AI. The assistant reads a prepared, pre-aggregated layer rather than raw tables, which is why the numbers it returns match your dashboards exactly.

Sources & destinations

Choose where your data, layers and models live.

BI.P.EYE runs the pipelines and writes your clean data, data layers and models to the destination you choose: your own Snowflake or Postgres, or an isolated BI.P.EYE warehouse that is yours alone. Keep it in your own warehouse, and you keep ownership of the storage, the access and the retention.

Your warehouse, your rules

Results are written to your own Snowflake or Postgres, next to the rest of your stack. You decide who can access it and how long it is kept.

A dedicated destination

Prefer us to host it? Run on an isolated warehouse, with your workloads and data separate from everyone else.

Private connections

Connect over AWS PrivateLink with no public endpoints. Data is encrypted in transit and at rest.

Terabytes of Parquet in S3?
Yes.CSV, JSON and Excel too.
Load it all into Snowflake?
Yes.Or into your own Postgres.
Latest file or the whole bucket?
Your choice.
Upsert or fresh write?
Both supported.
Refresh every 10 minutes?
Yes.Or on any schedule you set.
A ready-made MongoDB connector?
Yes.Pick a collection, no code.

Everything below is designed to cut analysis time

Connect, transform, visualize, and share. No code required. Built for speed and scale.

No-code pipelines

Create data pipelines with no code. Join, filter, and transform with clicks.

No-code charts

Build beautiful charts in minutes. Share or embed anywhere.

AI assistants

Live

Ask questions in plain language. Get queries and insights instantly.

Chat with your data

Natural language to SQL. Explain results. Iterate quickly.

Smart recommendations

Beta

Get suggestions for better visualizations based on your dataset.

Embed dashboards

Add dashboards to your app or website with a secure token.

Brand customization

Match fonts, colors, and layout to your product.

Scheduled jobs

Automate refresh with monitoring and progress tracking.

Permissions and caching

API keys, permissions and caching for fast, secure access.

Business intelligence, all in one platform

Most teams stitch together four tools: one to extract data, one to model it, one to visualize it, and one to embed the result. BI.P.EYE covers that whole chain. Data lands from your sources, moves through no-code transformations, gets pre-aggregated, and is served to dashboards, an AI assistant, or an iframe in your own product. Every step runs on a schedule you control, with progress tracking and alerts when a job fails.

Embedded analytics

Give your customers dashboards they'd pay for

Analytics is the feature your customers keep asking for and your roadmap keeps postponing. Drop BI.P.EYE dashboards straight into your product, in your branding, with every customer seeing only their own data. Weeks of work instead of quarters, and nothing to maintain.

Turn your product into a data product

Your customers stop exporting CSVs and start logging in more often. You ship a premium feature without hiring a data team.

  • • Interactive dashboards, not static reports
  • • Every customer sees only their own rows
  • • Your fonts, your colors, your product
  • • Live data, no exports, no stale numbers
Get started Book a walkthrough
1

Build the dashboard

Connect your data and design the dashboard with no code, the same way you would for your own team.

2

Decide who sees what

Set permissions per customer and per dataset, so each account is limited to its own rows automatically.

3

Drop it into your app

Paste one line into your product and it's live. Your engineers keep their sprint, your customers get analytics.

Developers get full control

Generate signed, row-level-secured dashboard links from your own backend, set expiry per link, and revoke access instantly. Everything is documented in the OpenAPI reference.

View API reference
Security architecture

Security is an architectural principle, not a checklist

Every change to the platform starts with one question: how could this be abused? Before we think about performance, UX or scale, we ask who should have access, how sensitive information moves, and whether the change weakens an existing control or opens a new attack vector. That applies to code, to infrastructure and to internal process, from day one and for as long as the product exists.

Connect your databases without exposing them

Connect your AWS databases and internal APIs over AWS PrivateLink. It creates a fully private connection between your AWS environment and ours, so your data never crosses the public internet, even across accounts and regions.

  • No public database endpoints
  • No inbound firewall rules for our static IPs
  • No internet exposure
  • Private VPC-to-VPC traffic, across AWS accounts and regions

Your database stays in your account, under your control. We connect to it, we don't take custody of it.

Cross-account, cross-region private connectivity over AWS PrivateLink.

Access control in layers

One permission model for people, API keys, embedded links and Claude, with row-level security (RLS) down to each customer's rows. Nothing is reachable until it is granted, and every request is checked again on the server, not only in the interface.

People
Roles, groups and sharing
  • Five roles, from Owner to Viewer, decide what each person can do across the organization: credentials, datasets, dashboards, API keys, users and billing.
  • Groups such as Finance let you give a whole team access at once.
  • On a single dashboard or dataset, raise a person, group or team above their role, or make it private to the people you choose.
Organization API keys
Each key has its own grants
  • A key is an identity of its own, not a person. A new key reaches nothing until you grant it models, datasets or dashboards.
  • Per key, decide whether it may turn on the AI chat in a link, whether it may create links without RLS, and the longest a link may live.
  • Turn a key on or off for each dashboard, separately from the people who can open it.
Row-level security (RLS)
Every customer sees only their rows
  • RLS rules per dataset on each embedded link, such as accountId is ac_123. Every chart, filter value, drill-down and chat answer is filtered by them.
  • Per link: a password, the IP addresses and ranges it may open from, and the sites allowed to frame it, on top of the dashboard's own lists.
  • A link never reaches more than the organization key that created it. Expire it in a day, or delete it and it stops working on the next request.
Claude and AI chat
AI sees what the asker may see
  • Choose which models are available in Claude: for the whole organization, a team or a single person.
  • Answers in an embedded dashboard use only the datasets that link shows, under its RLS rules.
  • Private datasets stay out of answers for anyone they are not shared with.

Every refused request is recorded with its reason, so an access review starts from evidence, not from guesses.

Embed through the OpenAPI, without exposing a key

Your server holds the organization key and asks BI.P.EYE for a short-lived dashboard link for each user. The browser only ever sees that link, already limited to the user's rows by row-level security (RLS).

  • The organization key never reaches the browser
  • One link per user or session, with its own expiry
  • RLS, password, IP and domain rules set in the same request
  • Update or revoke any link from your backend
View the OpenAPI reference
  1. 1. Your server calls POST /embed/link with the organization key.
  2. 2. BI.P.EYE returns a short-lived dashboard key and the URL to embed.
  3. 3. Your server sends your app the page with the embed URL, never the organization key.
  4. 4. The iframe in the user's browser loads the dashboard with the dashboard key.
The organization key stays on your server. The browser only gets a short-lived link, limited to the user's rows.
No public endpoints

Keep your databases off the public internet entirely.

Encrypted end to end

All data encrypted in transit and at rest.

Least privilege access

Authorized access only, scoped per user, key and dataset.

Built for compliance

Certified under ESOF AppSec ADA CASA, benchmarked against OWASP ASVS, and built to support SOC 2, ISO 27001 and PCI DSS programs.

Have a security review, a questionnaire or an architecture question? Email [email protected] and we will walk your team through it.

How to start

Test the full product in your own free trial

A step-by-step guide with a sample file of 1,000 Shopify orders. Run a pipeline, build a model, ask questions in a dashboard and in Claude, and see how embedding works. 30 days, full access, no credit card.

See the step-by-step guide Start free trial
  1. 1

    Add your data

    Upload a CSV or JSON file, or connect your app, database or warehouse.
  2. 2

    Run the pipeline

    Flatten, filter and select columns without code. Schedule it and write to your own warehouse.
  3. 3

    Build a model

    Connect datasets with relationships. One semantic layer for every dashboard and AI answer.
  4. 4

    Ask, then embed

    Chat with a dashboard or ask in Claude, then put the dashboard inside your product.

Start free. Scale when ready

Create an account and connect your first datasource without a credit card. Pricing is usage based, so a small project stays small on the bill. When your data volume grows, or you move dashboards into your own product, you move up a tier. If you need an isolated warehouse or a dedicated integration, email us and we will price it directly.

Create free account How to start
Use cases

Tell us your data problem. We'll solve it with you.

From a single Shopify store to a company running on twenty tools, we turn scattered data into decisions you can act on. Start with the case closest to yours, or send us the one that isn't here.

Shopify merchants

Find out where your store is losing money

Revenue looks fine, profit doesn't? The answer is hiding in returns, refunds, discounts and shipping. BI.P.EYE puts them all in one view so the leak stops being a mystery.

  • • Return and refund rate by product, variant and reason
  • • Connect carrier data: real shipping cost vs what you charge
  • • Trends and hidden insights, surfaced before the season ends
CEOs & COOs

One number the whole company agrees on

When every department brings its own spreadsheet, the meeting is about whose number is right, not what to do. Put everyone on one platform that syncs in real time, and challenge your data team and management layers with real data.

  • • CRM, finance and ops in one executive dashboard
  • • A single source of truth, refreshed on your schedule
  • • Built for mobile: your KPIs in your pocket, not in a deck
Marketing teams

Know which campaigns actually pay

Ad platforms grade their own homework. Pull spend, traffic and revenue into one place and see blended performance per channel, campaign and audience, without exporting a single CSV.

  • • Google Analytics, ad APIs and sales data joined automatically
  • • Spend next to revenue, not in separate tabs
  • • Ask the AI: "which campaign drove repeat customers?"
Ops & internal tools

Make your CMS and internal systems measurable

The most important data in your company often lives in tools nobody can report on: a CMS, an ERP, a homegrown admin panel. If it has a database or an API, BI.P.EYE can read it, model it and chart it.

  • • Any REST API, SQL or MongoDB database, or files in S3, no custom code
  • • Join internal data with Shopify, Stripe, sheets and more
  • • Scheduled refresh with alerts when a job fails
Every team

Surface the insights hiding in your systems

You can't ask a question you didn't think of. BI.P.EYE's AI agents read your prepared, pre-aggregated data and point at the things worth a second look: the segment that's quietly growing, the metric that drifted, the pattern nobody was watching.

  • • Insights on demand, in plain language
  • • Exact numbers: the AI reads a clean layer, not raw rows
  • • From "interesting" to a shared dashboard in minutes
SaaS & product teams

Ship analytics your clients will pay for

"Can we get a dashboard?" is your customers telling you what to build next. Embed interactive dashboards for sales, reports, usage and trends inside your product, in days instead of a quarter of engineering time.

  • • Row-level security: every client sees only their own rows
  • • Your fonts, your colors, your brand, our engine
  • • Iframe in minutes or a full Embed API from your backend

Don't see your use case? There are no limits here.

Ten orders a day or ten million rows an hour, one datasource or thirty: if your data lives somewhere, we can connect it, model it and put it to work. Tell us the problem, and we'll move fast, with numbers you can trust.

Get started Talk to us

Frequently asked questions

Can I ask questions about my company's data in Claude?

Yes. The BI.P.EYE MCP server connects Claude to your BI.P.EYE models, relationships and semantic layer. Claude queries the models, never your databases directly, and large results come back as a download link. Performance, caching, security and query efficiency are handled by BI.P.EYE. Add it from the Claude connectors directory, or follow the setup guide.

What is the semantic layer in BI.P.EYE?

A BI.P.EYE model maps tables from different sources to business entities such as customers, orders and products, defines the relationships between them, the metrics such as net revenue or return rate, and the access rules for each team or embedded customer. Dashboards, embedded analytics, chat and Claude all read the same model, so the numbers match everywhere. See how models are built.

Which datasources does BI.P.EYE support?

BI.P.EYE connects to a wide range of sources, and we add more all the time:

  • Databases and warehouses: Postgres, MySQL, SQL Server (MSSQL), MongoDB, Snowflake, BigQuery, Redshift
  • Files in Amazon S3: CSV, JSON, Excel and Parquet
  • eCommerce and SaaS: Shopify, Stripe, Deel, Bubble, Mixpanel, SkuSavvy, currencyapi.com
  • Google: Analytics, Drive, Sheets
  • Cloud: AWS
  • Any REST or HTTP API: pagination, authentication, rate limits, dependent requests, failure recovery and nested JSON are handled for you

Don't see your source? We build dedicated integrations on request. Reach out at [email protected].

Can I embed dashboards in my product?

Yes, in two ways. For a quick setup, generate a dedicated API key in the dashboard and drop our iframe snippet into your product. For full control, use the Embed API to generate signed embed links from your backend.

Access is protected by several permission layers. You can scope per user and per end-user, by API key, by IP address, by allowed domain, and by dataset-level rules. Each API key can also carry per-dataset row-level filters, so every customer sees only their own rows.

How does AI help?

AI works at every step, from raw data to the answer:

  • Ask your data. Chat with a dashboard in plain language, in English or Hebrew, and get back a number, a chart or a table.
  • Build and fix charts. Describe the chart you want, or ask the chart to change, explain itself or point out what stands out.
  • Write SQL. Turn a question into SQL that knows your schema, then review, edit and run it.
  • Model your data. AI maps tables to business entities, metrics and relationships, and suggests how hundreds of tables connect.
  • Collect and prepare data. Agents pull data from APIs, analyze database structures and improve ETL pipelines.
  • Give your customers an AI analyst. Turn on the chat in an embedded dashboard, and each customer can ask about their own data only.
  • Use Claude. Connect Claude to your models and ask questions there, without leaving the conversation.

Every answer shows the decisions and assumptions behind it, such as which dataset and date column were used, so you can check how a number was calculated. And the AI only sees what the person asking may see: the same roles, sharing and row-level security (RLS) apply to it.

Why not just send my data to an LLM and get answers?

It comes down to money, performance and security. You cannot hand a language model millions of rows for every question. It is slow, and the bill climbs with each query. Shipping raw data to a third-party model is also a security and privacy risk. Models tend to make mistakes when they compute over raw data on the fly.

BI.P.EYE does the hard work before you ask. Your data is collected, validated, structured and pre-aggregated by a real BI engine. The AI then sits on top of that clean layer. You get exact numbers, answers in seconds, and a fraction of the cost, without exposing raw data to the model.

What about security?

Every change to the platform starts with one question: how could this be abused? All data is encrypted in transit and at rest, and the infrastructure runs on AWS, MongoDB Atlas and Cloudflare.

  • Access control in layers. Roles, groups and per-dashboard or per-dataset sharing for people. Organization API keys with their own grants. Nothing is reachable until it is granted.
  • Row-level security (RLS). Each embedded link carries its own rules per dataset, so every customer sees only their own rows, in charts, filters, drill-downs and AI answers.
  • Secure embedding. Your server creates short-lived links through the OpenAPI, so the organization key never reaches the browser. Each link can also require a password and be limited to IP addresses and allowed domains.
  • Private connectivity. Connect AWS databases and internal APIs over AWS PrivateLink, with no public endpoints.
  • Audit. Every refused request is recorded with its reason.

Our application security practices are certified under ESOF AppSec ADA CASA, an independent assessment by TAC Security. CASA is benchmarked against the OWASP Application Security Verification Standard (ASVS), the most widely adopted standard for application security requirements, and tests controls across fourteen categories including authentication, session management, access control and cryptography. The framework requires annual re-assessment, so it reflects the platform as it is today, not as it was at launch. We monitor, update and improve our systems continuously. See the security architecture for details.

Can I connect my database without exposing it to the internet?

Yes. You can connect AWS databases and internal APIs over AWS PrivateLink, which creates a fully private connection between your AWS environment and ours. That means no public database endpoints, no inbound firewall rules for our static IPs, no internet exposure, and private VPC-to-VPC traffic even across AWS accounts and regions. See the security architecture for the full picture.

Can you build a full BI project for us?

Yes. Beyond the self-serve platform, our team can design and build a complete BI project for you: data modeling, pipelines, dashboards and embedded analytics. Reach out at [email protected] to scope it.

Where does my transformed data land?

You choose the destination. ETL results can be written to the BI.P.EYE warehouse or to your own Postgres or Snowflake warehouse, so your clean, transformed data lives where the rest of your stack can use it.

How granular are permissions?

Each person has one role across the organization (Owner, Modifier, BI Engineer, Data Analyst or Viewer), and you can raise a person, group or team above it on a single dashboard or dataset, or make that dashboard or dataset private. Organization API keys have their own grants: models, datasets or dashboards. Each embedded link carries its own row-level filters, and can be limited by password, IP address and allowed domain. See the security architecture.

Can you add an integration that isn't listed?

Yes. Our API connector already covers most HTTP APIs. If you need a dedicated integration built, talk to us at [email protected] and we will add it.

Can my data run in an isolated warehouse?

Yes. You can run on an isolated warehouse for faster queries and stronger data isolation, keeping your workloads and data separate from everything else.

How fast is support?

Fast. Typical response time is a few hours at most. You can reach us at [email protected] any time.

Do you take feature requests?

Always. We build quickly and we like feedback. Tell us what you need at [email protected]. It shapes the roadmap.

Still have a question? Email [email protected] and we will get back to you within a few hours.

Contact

Questions about the platform, embedded analytics, or a custom BI project? Email [email protected]. We reply within a few hours on business days. You can also reach the team on LinkedIn.