Your data.
Your questions.
A clearer picture.

Connect business data, ask questions in plain language, build dashboards, write SQL, and work in Python notebooks. Agent Foundry lets you turn recurring analysis into monitored workflows, while sharing and administration controls help teams work together.

No coding is required to start a conversation or build dashboard queries by choosing tables and columns. SQL Work Bench and Python notebooks are available for deeper hands-on analysis. Connecting data sources may require help from someone familiar with your systems.

Sign in, verify your email, and create an organisation or join through an invitation. Your organisation manages its plan, departments, and members. Once you have access to a department and its data sources, you can start creating and sharing work.

No. The revenue conversation uses sample data, and the finance, marketing, and operations images are conceptual illustrations. Inside your workspace, analyses use the sources you connect and have permission to access.

Supported database connectors include PostgreSQL, MySQL, BigQuery, and MongoDB, alongside CSV and Excel uploads. Excel worksheets become individual queryable tables. Connected SQL sources and imported spreadsheets can be queried together, subject to your access permissions.

Yes. Attach documents and files to chats, organize files in knowledge-base folders, and connect vector stores such as Pinecone, Weaviate, or ChromaDB for semantic retrieval. You choose which available sources a conversation can use.

Combine charts, tables, KPIs, and notes. Build queries visually or write them directly, map fields to visualizations, and connect shared filters across widgets. The dashboard AI assistant helps in context, and version history lets you save named versions and restore earlier layouts.

Use SQL Work Bench to write queries, reference source tables, inspect results, and save sessions. Python notebooks run in isolated Jupyter sessions and support .ipynb import and export. Notebook execution requires the notebook runtime to be enabled for your deployment.

The Beta modelling editor lets you describe tables and relationships, including reviewing AI-suggested joins. Schema drift tracks structural changes separately, with alerts, snapshots, and before-and-after comparisons that owners and editors can acknowledge or resolve.

Yes. Agent Foundry supports scheduled runs, data-change triggers, webhooks, and manual runs. Configure the agent’s model, tools, data access, and limits, then review its activity and outputs. Configured alerts can be delivered in-app, by email, or to Slack.

Yes. Proposed actions appear in the agent’s outputs for approval or rejection. Governed actions such as external API calls, configuration changes, and budget changes require human approval. External calls also need an administrator-approved destination; agent access is limited to its configured tools and sources.

You can review and edit your personal memory document, inspect its history, and restore an earlier version. Chats and agents keep separate context, and memory can be switched off for a chat or agent. Preferences can be tied to particular data sources or dashboards instead of applied everywhere.

Yes. Administrators enable available providers, then members connect their own accounts. Connections are personal, and agents use tools approved for them. Administrators can also register custom remote MCP servers on approved hosts. Providers marked coming soon are not yet connectable.

Normal resource listings are private by default: people see what they own or what has been shared with them. Share with a person, department, or organisation using Viewer or Editor access. Administrators have additional access within their role; supported admin-view access is acknowledged and audited.

External sharing is available when your organisation’s administrator permits it. Supported external links are revocable and can have an expiry. Review the audience and permissions before sharing: a link can be forwarded, so it should only include information you intend to share externally.

Administrators can manage approved models and tools, agent query and spend limits, data policies, sensitive-field handling, and security scans. Audit logs show recorded activity. Usage and billing views track consumption and departmental credit allocations; some controls depend on role, plan, and configuration.

The dedicated Data Pipelines workspace is still coming soon. For recurring analysis today, use Agent Foundry schedules and triggers. Data-source sync schedules are also available for supported connectors; CSV and Excel sources refresh when uploaded or manually re-uploaded.