Bonnard

We tested Bonnard against some of the most popular BI tools.

Correct answers

+26% accuracy

vs the average of six BI MCP servers, same 20 questions

Time per answer

42% faster

22s vs 38s for the next fastest

Tool output per answer

13k tokens

Next lowest: 25k

Against six other BI products we tested

20 business questions, 3 tries each, on a dbt-built warehouse with metric definitions loaded.

Method on request

Self-serve

Let users help themselves reliably

Users can build, save and share everything from a quick chart to a full dashboard just by asking in plain language.

The request never reaches you

Business users answer their own questions in the client they already have open, so the ask stops arriving in your queue.

Scoped to the app you publish

Pick each app’s tables and tools. Your database grants set what it can read.

Unlimited users on every tier

Every plan carries unlimited users, apps and database connections, so opening this up to the whole company adds no seats.

What your users get

A BI toolkit built for AI

Let users query and share without leaving their AI client.

  • Queries and visualisations. Query and visualise your data directly from the AI client you already use.
  • Workflow tools. Bonnard slots into your existing workflow, letting you create, view, download and share your data without leaving your AI client.
  • Dashboards. Describe a dashboard, then save and share it by link.
  • Exports. Save any result as CSV or PNG.
  • Edit in Bonnard. Edit saved dashboards in Bonnard’s web app.
  • Desktop and mobile. Query and view your data on desktop or on your phone, wherever you’re working from.

Data sources

Context you already wrote

The agent reads the descriptions and keys already in your database.

Connected read-only

PostgreSQL, BigQuery, Redshift, ClickHouse, SQL Server and more. One per app.

A sample database to start

Try charts on a hosted shop database in one click.

Credentials encrypted at rest

Bonnard stores connection credentials AES-256-GCM encrypted and keeps API keys only as SHA-256 hashes.

Setup

Four steps,
no engineering

  1. Connect your data

    Databases, warehouses and PostHog, connected read-only.

    • PostgreSQL
    • BigQuery
    • Redshift
    • ClickHouse
    • MotherDuck
    • MySQL
    • MariaDB
    • SQL Server
    • Azure SQL
    • PostHog
  2. Create an app

    Choose the tables it can see and the tools it offers: charts, dashboards, exports.

  3. Choose the audience

    Your team, all your customers, or the ones you name.

  4. Share the link

    Users connect from the AI client they already use.

Monitoring

From call log to data model

Failed queries show where your data model needs work.

LogsLast 30 days
  • 1m agosql_queryok61ms

    SQL SELECT week, channel, SUM(revenue) FROM orders GROUP BY week, channel

  • 1m agorender_chartok38ms
  • 2m agoexplore_schemaok40ms
  • 4m agosql_queryerror30,000ms
  • 6m agosql_querydenied22ms
  • 7m agorender_dashboardok112ms
  • 9m agoexplore_schemaok44ms

Every call, with who and when

Bonnard records every tool call with who ran it, when, which tools ran and what failed, per app and per customer.

The calls that failed

Failed and denied calls sit in the same log as the answered ones, so a gap in the model reads as a pattern.

Answer who queried what

Every query, per app and per customer, with no logging layer to build.

Security

Permissions that travel with your data

Each query runs as one customer, under the tenancy you already use.

Tenancy set per database

Row-level security, or a schema, database or dataset per tenant.

Your IdP decides the tenant

A sign-in through your identity provider maps to the right tenant. Scale and Enterprise plans.

An audience per app

Anyone outside it is refused, and each refusal is logged.

Put your data team back on the work that matters.

Free while you find out whether it sticks. No seats, no minimums, no card to start.