New: give your agents governed data they can query and write to, no warehouse to run.Get started
Nightshift

A data platform built for composition

Analytics at the speed of thought.

Connect any source and let your people and your agents query it live, no warehouse to run.

Want to look first? Take the product tour·Add Nightshift over MCP

Works with any agent, over MCP

Claude CodeCursorGitHub CopilotWindsurfOpenAI CodexLangGraph+ your in-house agents

Everything you need to put your data to work.

Connect a source, query it like a warehouse, serve any consumer, and build real apps your agents and people can write to. One layer does it all, governed by identity so it is safe by default.

Built for composition.

Compose every source into one governed layer, then serve it live to any app, agent, or dashboard.

  • Databases, warehouses, SaaS APIs, and your own services, side by side
  • One layer feeds every consumer: apps, agents, and dashboards
  • Composed across systems in real time, with no pipelines to maintain
Explore composition
Composition4 sources · 3 consumers
SnowflakeBigQueryPostgresSAPData appapp.tsxAgentover MCPDashboardanalytics
Composed across systems, in real time

Beautiful apps on live data.

Build hosted, shareable apps on your governed data in minutes. Real-time charts and tables your team and your agents actually use, with no front-end project to staff.

  • Hosted and shareable, secure by default
  • Real-time on live data or a pinned snapshot
  • One governed endpoint behind every app
See it in action
Revenue by regionLiveshared

Net revenue · 24h

€4.28M

EMEAAmericas

Customers

1,284

EMEA share

38%

Avg contract

€61k
A data app on one governed endpoint · hosted, shareable, recorded

Write back, not just read.

That app has a real backend, and agents and people write to it. Every write lands in a managed, versioned Nightshift table, and never touches your system of record.

  • Compose columns from every source, then write your own on top
  • Agents and people write to governed Nightshift tables and views
  • Every write versioned and queryable as of any point in time, system of record untouched
How write-back works
ns.collections_worklistv7
invoiceNetSuitebalanceNetSuiteaccountSalesforcepriorityNightshiftactionNightshift
INV-2231$48,200AcmeHighcall
INV-2218$12,400GlobexMedemail
INV-2205$91,750InitechHighescalate
INV-2199$7,300UmbrellaLowemail
INV-2184$26,500SoylentHighcall
NetSuite and Salesforce stay read-only

Fiber, the typed React API.

Nightshift Fiber turns a query into a typed result and binds it straight to your UI. Build on your governed data in a few lines of React.

  • Typed queries from your SQL, params bound server-side
  • Bind results to your own components, tables, and charts
  • Credentials stay server-side, governed by identity
Explore Fiber
app.tsx@nightshift-sdk/fiber
import { useQuery, LineChart } from '@nightshift-sdk/fiber'

const rows = useQuery(revenueByMonth, { region })
return <LineChart query={rows} x="month" y="revenue" />
$ npm i @nightshift-sdk/fiber

Edge speed, no cluster.

All the speed of a warehouse: no warehouse to stand up, no per-query compute bill. Analytics run at the edge of your source.

  • Compute runs at the edge of your source, with no cluster to stand up
  • Run as many iterative queries as you need, with no warehouse compute to pay for
  • Heavy workloads run on an isolated path, clear of the database your people depend on
How the engine works
arr_by_regionat the edge0.41 s · no cluster
at sourcescanned and aggregated9.8M rows
over the wirereturned to the agent312 rows
The filter and the sum ran where the data lives. Only the result crossed the wire, with no warehouse in between.

A real query engine

The depth real analysis needs.

A full-featured edge analytics engine, built for analytics at the speed of thought.

$0 per-query

Analytics pushed to the edge the moment a query lands. No warehouse to provision, no cluster sitting idle, and no per-query meter running.

Sandboxed

Every query runs in its own sandbox, so its cost and load stay isolated from your bill and from other workloads.

Views

Define a view once and everyone queries it like a table, with the heavy joins and aggregates precomputed for them.

Snapshots

Pin a point-in-time view of any source for reproducible analysis, right alongside live reads.

Time travel

Query the data as it was at any moment, so an agent or an auditor can see exactly what changed.

Audit log

Every query on the record: who asked, what policy did, and what came back, queryable back to day one.

Query

Add Nightshift to whatever agent you already use.

Nightshift is an MCP server. Drop it into any agent and the governed catalog shows up as a tool. Pick your agent for the exact setup.

Claude Code

Add the server with one command.

terminal
claude mcp add nightshift \
  --url https://mcp.nightshift.sh

The catalog appears as a tool. Every request it makes is shaped by your policies before a row is returned.

The bill

The cheapest query is the one you don’t pay a warehouse for.

Warehouse spend is becoming one of the biggest lines on a cloud bill, and giving agents unrestricted access to your warehouse only inflates it. Nightshift pushes the analytics to the edge of your source, so iterative work and rogue queries never run up a metered compute bill.

Off your invoice

  • Warehouse compute, scaled up or sitting idle
  • A second copy of your data to store and secure
  • Pipelines to move rows you already have
What’s left is the request you actually ran, against data that stayed where it lives.

Data your team and your agents can be trusted with.

Connect a source and point your team or any agent at warehouse-grade analytics, no warehouse required. Governed by identity. Start free in minutes.