August 19, 2026

New Additions

Agent Cost Dashboard
A new dashboard is available detailing cost centers of agent usage across users and agents with anomaly detection.

bigAI Chat is now generally available
Ask bigAI about your data in plain language and get answers — root causes, lineage, downstream impact, monitor history, and AI-agent cost — that link straight to the issues, tables, and lineage they reference.

Business Glossary in beta
Define governed business terms with owners, stewards, statuses, and synonyms, and link them to the catalog assets they describe.

Multi-factor authentication
Add a second factor to password-based sign-in, with authenticator-app enrollment, recovery codes, and an option to require it across your organization.

Claude Code monitoring in Agent Trust
Ingest Claude Code telemetry directly into Agent Trust to see each conversation's turns, tool calls, and the data it accessed.

Added

bigAI Chat

Ask what caused an issue and get the upstream tables and columns where it originates, plus the issues on them
Ask for any table or column's lineage — what feeds it and what depends on it
Ask which tables, reports, and owners are downstream of an issue, and which issues are affecting a given BI report or dashboard
Ask about a monitor's run history — past values, trends, and why a metric, dbt test, or SQL/join rule failed or alerted
Ask which users are driving agent conversation cost, and who has accessed a given table or column
Ask about the agent, conversation, or issue on the page you're viewing
Ask about a custom rule's SQL, thresholds, and schedule
Create custom SQL rules, run metrics, and set or clear a table's row creation time — all with approval
Replies link out to the lineage graph, issues, tables, columns, schemas, sources, collections, deltas, and AI agents

Business Glossary

Glossary list page with search, status and owner filters, and inline status editing
Term detail page with linked assets, owner and steward editing, synonyms, and a change timeline
Link terms to catalog assets, and see a term's data quality and classification at a glance
Linked glossary terms shown on catalog detail pages
New "Edit Glossary" role permission for managing terms

Security

Multi-factor authentication for password sign-in, with authenticator-app enrollment and single-use recovery codes; admins can enable or require it per organization
New "Allow password authentication for the API" setting to require personal API keys, closing the gap left when passwords still reach the API directly

Agent Trust

Native ingestion of Claude Code telemetry as conversations, turns, and data-access records
Governance score on the AI agent registry, and a Governance column on the data-access list, showing the percentage of accessed assets that are certified
Monitoring tab on the conversation detail page showing metrics scoped to that conversation
Add monitoring or configure scans directly from a conversation, with source picking and pre-selected tables
"Sync agent prompts and responses" setting to control whether raw prompt and response text is stored

Metrics & Rules

Select all custom rules matching the current filters and run them in one batch
Grouped metrics can alert and open an issue when a group disappears, in Bigconfig
Custom lineage node types can opt into impact analysis and appear in an issue's root cause

Platform

Command palette shortcuts for Setup, Home, Search, Lineage, Data Classification, Agent Trust, and most settings pages
Bulk endpoint for setting catalog attributes across many entities in one request

Updated

bigAI Chat

Now generally available — the "Beta" label has been removed
Available to everyone and enabled by default after moving through public preview
Approval cards describe changes in plain language, with human-readable rule summaries instead of raw JSON
Asking about an issue now includes Bigeye's own AI analysis, recommended resolution steps, and how similar issues were resolved
Respects workspace edit permissions — users without edit access are told what's required instead of having changes made

Agent Trust

Claude Code telemetry rolls up under a single Claude Code agent, with the person who ran each session shown as the conversation's user
Tool-call sources show real logos and human-readable labels (Linear, Google Drive, Datadog, Notion, SharePoint, Bash, Read/Write/Edit file)
Access-detail groups are collapsible and sorted by decision severity
Accessed assets link to their catalog page

Catalog

New Snowflake connections default to key-pair authentication, as Snowflake phases out password-based accounts; "Username and password (legacy)" remains selectable, and existing sources are unchanged

Lineage

Lineage search uses indexed, per-word matching for faster, broader results
"Show agent nodes" toggle to filter AI agent nodes from the graph
Topology graph layout is now on for everyone
More reliable syncing for very large metadata repositories, with lower memory use and resumable progress

Metrics

More reliable automatic threshold model training, with training data encrypted and compressed in transit

Fixed

bigAI Chat

A chat that fails to load now shows an inline retry instead of a blank panel
Assigning issues to a teammate by name now works
Creating a metric no longer takes several attempts and repeated approvals
Clearer messaging when an item can't be found, instead of a permissions error

Agent Trust

Claude Code conversations now report full token usage and cost, including prompt-cache reads and writes that were previously discarded
BigQuery and Snowflake Intelligence agent SQL is parsed into the real tables and columns touched, linked to the catalog
MCP and tool calls (Slack, Notion, Google Calendar, WebSearch, subagent tasks, Bash commands) show what was actually accessed instead of an opaque identifier
Empty duplicate conversations are no longer created for sessions that did no work
The Monitoring tab on the conversation detail page no longer appears empty
The Agent cost dashboard now shows actual cost instead of "No cost data" and $0.00

Metrics

BigQuery metrics with a partition filter no longer fail to run or validate
Databricks freshness and volume metrics now detect updates made via multi-statement transactions
Freshness and volume metrics no longer fail on Redshift Serverless sources
No more false anomaly alerts on Volume metrics caused by future-dated timestamps
Anomalous values marked as alerts no longer widen autothreshold calculations
"Hours since latest value" no longer appears as duplicate metric types in the deploy wizard

Lineage

Lineage search no longer returns few or no results
Table-level lineage flows are no longer dropped during sync

Deltas

Deltas no longer show as "currently running" or "last run failed" indefinitely after completing

Notifications

Notification delivery no longer times out for grouped metrics with very large runs

Accessibility

Improved VoiceOver arrow-key navigation in the status, priority, tag, and owner dropdowns

Copy-to-clipboard buttons no longer submit the form they sit within