Back

Pricing Customers
Home Blog

What's new in Toucan: May 2026 product update

icon-briefcase-blue

What's new in Toucan: May 2026 product update

Résumer cet article avec :

Good embedded analytics is not just charts. It is governed metrics, transparent AI, and the platform controls that let you ship analytics to customers without losing sleep over consistency or cost.

Six updates are shipping now. They touch the chart editor, exports, the AI agent's transparency, the filter bar, database connections, and AI spend. Here is what changed, and why it matters if you are embedding Toucan in your product.

Charts

Chart operations, now in the semantic layer

Charts
Edit column · Usage Count Expression
Basic Expression
Expression *
SUM(count)/COUNT(count)
SUM ( 01/10 count ) ÷ COUNT ( 01/10 count )
 
Additional information Metric
Ordering

The chart editor now runs on the semantic layer. That unlocks more calculations, custom formulas, and expressions directly where you build the chart, with no detour through a separate modeling step.

The point is not just more math. Because expressions resolve through the semantic layer, the numbers stay governed and consistent across every dashboard and every tenant. One definition, used everywhere.

Chart operations are rolling out behind a feature flag. To switch them on in your environment, set localStorage.setItem("ENABLE_SEMANTIC_CHARTS", "true") in your browser console.

Dashboard

Export any chart to CSV or XLS

Dashboard
Revenue by connector
postgresql
bigquery
snowflake68
clickhouse58
05001,0001,5002,0002,500
CSV
XLS
↓ Download
↓ Delete

You can now download the data behind any chart, straight from the dashboard, as CSV or XLS.

It is a small thing that closes a real gap: the people who need the numbers are not always the people logged into Toucan. Now your users can hand off a file without asking you to pull an export for them.

Filters, compact by default

Dashboard
Dashboard · Filters Compact
User Type
Select…
Date d'utilisation
Jun 22 – Jun 22 2026
Use Case
Select options…

The filter bar is now more compact and visually consistent with the rest of the interface. Less vertical space, same control.

On an embedded dashboard, every pixel you give back to the actual data is a pixel that makes the experience feel like part of your product, not a tool bolted on the side.

AI agent

Reasoning you can actually follow

AI Agent
AI Agent Reasoning visible
What drove the churn spike in Q2?
Reasoning chain
 
Understood intent+0.1s
Detected churn analysis · time scope Q2 2026
Queried semantic layer+0.4s
Fetched churn_rate, MRR, cohort segments
semantic_layer.query
Ran trend analysis+0.9s
Detected +4.2pt spike in May, correlated with plan changes
trend_analysis.run
Generated answer+1.2s
Structured response with supporting chart

The reasoning chain the agent shows after each answer is now structured and chronological. Every step and every tool it called, in order, with timing.

For an AI feature you are putting in front of customers, this is the difference between trust and a black box. When a user can see that the agent queried the semantic layer, ran a trend analysis, then wrote the answer, they believe the result. So do your support and sales teams when a customer asks how it works.

Platform

Database errors that explain themselves

Platform
Connection settings
Host
{ }  db.acme-corp.io
Port
{ }  28560
Database
{ }  production_db
SSL
Connection refused
The database server is not reachable at the host and port you provided. Check that the server is running, the port is correct, and our IP is whitelisted.
^ Technical details
got error Connection refused (os error 61)
for socket_addr db.acme-corp.io:28560
Test connection Connect →

When a database connection fails, users now get a clear title, a plain-language explanation, and copyable technical details. No more deciphering a raw stack trace.

Connection issues are the most common snag when onboarding a new data source. A readable error turns a support ticket into a self-serve fix, which matters when your own customers are the ones connecting their databases.

AI credit limits per user

Platform
Settings · AI usage Admin
Monthly AI credits limit
Cap the number of AI credits each internal and embedded user can consume per calendar month. The limit resets on the first day of each calendar month and is independent of your billing cycle.
 
Enable monthly AI credits limit per user
Monthly credits limit per user *
1000
Save

Org admins can now set a monthly AI credit cap per user. Full control over usage, with no billing surprises at the end of the month.

If you are embedding AI analytics across a base of customers, this is the cost guardrail you need: predictable spend per seat, set by you, independent of your billing cycle.

Ship analytics your customers actually use

All six updates are live. If you already embed Toucan, they are in your environment now.

If you are still weighing build versus buy for in-product analytics, here is the short version: Toucan is AI-native embedded analytics for SaaS products, multi-tenant by design, with your first dashboard shipping in days rather than quarters.

Try for free Book a demo