Ask once. Agents do the rest.
AI analytics in plain English. Ask a question, get an answer you can check, and let 20+ agents connect the source, write the query, build the dashboard and deliver the report. They work inside the data layer, against your schema and your governed definitions. Your model or ours.
In-app · over MCP · embedded in your product · in Slack and Teams
Search-based analytics answers the question. An agent carries it through.
From question to delivered answer.
You ask once. The orchestrator works out what is needed, picks the right agents and chains them with shared context. You never choose an agent from a menu.
Plain English
“Create a dashboard showing email campaign metrics for Q1.” No query language, no chart picker.
Intent and sources
Detects what you meant, selects the data sources and picks the agents to run in what order.
Query and build
Agents connect to the source, run the query and assemble the dashboard against the real schema.
Where you work
“Send this to Slack every Monday.” Reports go to email, Slack, Teams or a webhook.
20+ agents. Each does one job.
The orchestrator decides which ones a request needs and chains them with shared context.
Agent Orchestrator
Takes your input, detects intent, selects and chains the agents, and maintains shared context between them. You describe what you want. It works out how.
Connect Datasource
Walks you through connecting a new source in conversation rather than a settings screen.
Query Your Data
Natural language to SQL, NoSQL or API calls, auto-detecting the right source.
Data Analyst
Answers questions using datasets and documents. Produces analysis, not just a query.
Create Dashboard
A full dashboard from one prompt, with layout, charts and filters chosen automatically.
Create Widget
Builds individual visualizations with chart type and field mapping selected for the data.
Document AI
Queries PDFs, Word files and spreadsheets alongside live database data.
Send Report
Delivers dashboards and widgets to email, Slack, Teams or a webhook, on a schedule.
Manage Alerts
Creates and configures data alerts that fire on the conditions you define.
Get Recommendations
Insights on any widget: anomaly detection, trend surfacing and pattern discovery.
Inside the data layer.
Most tools wrap an API around a model and let it infer structure from the outside. Knowi's agents read the schema, field types and relationships directly, which is why one call builds a dashboard instead of ten.
- Direct access to schema, field types and relationships
- Chart type and field mapping chosen from the real data, not inferred
- 70+ sources with native NoSQL and SQL support
- Cross-source joins without ETL or data movement
Knowi MCP, in Claude and other agents.
Knowi ships as an MCP server with full orchestrator access. The agents run where your team already works, under the same governed datasets and permissions.
- Query any connected data source
- Create and modify dashboards
- Trigger scheduled reports
- Full agent orchestration over MCP
You: Revenue by region last quarter
Claude: North America $2.4M, Europe $1.8M...
You: Create a dashboard with that
Claude: Dashboard created.
app.knowi.com/d/85030
Private AI. Your model or ours.
Knowi runs its own models, inference and vector search on Knowi's infrastructure, or inside your environment on-premises, so queries and results are not handed to a third-party LLM. Or bring your own Azure OpenAI or Claude key. Choose per feature.
- Built-in models and vector search, no external call required
- On-premise deployment of the full platform and your own LLM, via Docker or Kubernetes
- Choose Knowi AI, OpenAI or Claude per feature, with no lock-in
- SOC 2 Type II, and deployable for HIPAA-regulated workloads
Knowi agents, inside your product.
Call the agents from your own copilot, chat UI or internal tool to query data, build dashboards and deliver reports. Your users never see Knowi.
- Your interface, Knowi's agents behind it
- Full API access to every agent capability
- Multi-tenant isolation, row-level security, SSO, on-prem
- Internal copilots, customer-facing analytics, vertical AI assistants
One prompt, per team.
Same orchestrator, different sources and different chains.
- connect datasource links Salesforce
- query agent pulls pipeline data
- create dashboard builds the view
- send report delivers to Slack weekly
- connect datasource links Klaviyo and Shopify
- query agent joins campaign data to revenue
- create widget builds the comparison chart
- get recommendations surfaces top performers
- connect datasource links Stripe and QuickBooks
- create dashboard builds the revenue view
- manage alerts sets the MRR threshold
- send report delivers the monthly summary
- connect datasource links product DB and Salesforce
- data analyst calculates usage trends
- get recommendations flags at-risk accounts
- manage alerts triggers weekly churn warnings
- connect datasource links Jira and PagerDuty
- query agent joins deploy and incident data
- create dashboard builds DORA metrics
- dashboard summary writes the weekly narrative
- connect datasource pulls from every connected source
- data analyst compiles cross-department KPIs
- dashboard summary writes the narrative
- send report delivers to Slack on schedule
The same agents, wherever the work happens.
Conversational Knowi
Chat history, follow-ups, file uploads and multi-source querying inside the platform.
From Claude or any client
Full agent orchestration from any MCP-compatible tool your team already uses.
Your product, our agents
Your own copilot or app powered by Knowi agents via API. White-label and multi-tenant.
In the channel
Ask questions and get reports delivered directly where the conversation already happens.
Questions about agentic BI.
What is natural language BI?
Natural language BI lets you ask a question of your data in plain English instead of writing a query. Knowi resolves the question against the real schema and governed business terms, then returns an answer, a chart or a full dashboard rather than just a result set.
What is search-based analytics?
Search-based analytics is the pattern of querying data by typing a question the way you would type a search, rather than building a report. Knowi extends it: the question does not just return a result, it triggers agents that build the visualization, schedule the report and set the alert.
What is agentic BI?
Agentic BI uses specialized AI agents to handle the full analytics workflow: querying data, building visualizations, creating dashboards, detecting anomalies and delivering reports. Instead of a single chatbot, an orchestrator routes each request to the right agents and chains them together.
How are Knowi's agents different from a BI chatbot?
A chatbot answers one question at a time with no memory. Knowi's orchestrator chains up to five specialized agents per request with shared context. The Query Agent writes the query, the Dashboard Agent builds the layout, the Send Report agent delivers it to Slack. They work as one sequence.
What is MCP and how does Knowi use it?
MCP, the Model Context Protocol, lets AI tools such as Claude connect to external services. Knowi ships as an MCP server, so you can query databases, build dashboards and trigger reports from any MCP-compatible client.
Where does the AI run?
Knowi runs its own AI: models, inference and vector search on Knowi's infrastructure. You can also choose OpenAI or Claude per feature, or run the full platform and your own LLM locally via Docker or Kubernetes.
Can I embed agents into my own product?
Yes. The AI agent widget can be white-labeled and embedded into your SaaS product with multi-tenant isolation, row-level security, SSO and full API control. Your customers never see Knowi branding.
What data sources do agents support?
70+ native connectors: SQL including PostgreSQL, MySQL, SQL Server, Snowflake, BigQuery and ClickHouse; NoSQL including MongoDB, Elasticsearch, Cassandra and Couchbase; REST APIs, cloud services, and documents such as PDFs, Word, Excel and Google Sheets.
Do I need to build a semantic layer or data model first?
No. Knowi agents query your sources directly, with no LookML and no cube definitions to author before the first question. If you want governed business terms enforced across every agent and dashboard, Knowi's semantic layer is there to add on top rather than a prerequisite in front.
How many agents are there?
More than 20 today, covering connection, querying, analysis, dashboard and widget creation, document Q&A, alerting, delivery and recommendations, with more shipping regularly.
What is AI analytics?
AI analytics is analytics where the model does the work rather than describing it: it interprets the question, writes the text to SQL translation, picks the visualization and explains the result. In Knowi that work is split across more than 20 agents, and each step stays inspectable, so you can see the query behind any answer.
Can AI build a dashboard from a question?
Yes. Ask for what you want to see and the Create Dashboard agent generates the queries, selects chart types and arranges the widgets. The result is an ordinary Knowi dashboard you can edit, share, schedule or embed.
What is an AI data analyst?
An AI data analyst answers data questions end to end rather than returning a single chart: it clarifies the question, finds the right sources, runs the query, checks the result and writes the summary. Knowi's agents cover those steps against live SQL, NoSQL, API and document data.
One prompt. The whole workflow.
See the orchestrator chain agents against your own data.
20+ agents · 70+ sources · in-app, MCP, embedded, Slack