The data and agent platform for regulated work

Connect your systems.Model your business.Put agents to work, measured.

NSTACK builds one model of your business from the systems you already run, then puts agents to work on it. You set the limits. Every run shows what it was worth.

Talk to us See how it works

Live todayAn SEC-registered investment adviser runs NSTACK on its own private instance.

Works with your data and your agents' tools

  • Snowflake
  • Model Context Protocol
  • Databricks
  • Claude
  • PostgreSQL
  • Gemini
  • Salesforce
  • OpenAI
  • Amazon S3
  • Oracle
  • Google Cloud Storage
  • Any HTTP API

01·The gap

Your business already has the answers. They're scattered across a dozen systems.

Customers live in the CRM. Transactions in the core system, history in the warehouse, the rules in a few people's heads. Point a language model at that and you get a clever paragraph. You need work you can approve, repeat, and measure.

02·The three layers

NSTACK puts it back together, one layer at a time.

Your systems stay put.

  1. 03Experience
  2. 02Model
  3. 01Systems

01 · Systems

Connect what you already run.

Warehouses, databases, CRMs, and file stores connect as they are. Keys go in once, never shown again. No API? Show Scan the task once, and an agent repeats it in a real browser.

The Scan run bar: Safe mode and Autonomous, Publish to Studio, and Run again. The Scan run bar: Safe mode and Autonomous, Publish to Studio, and Run again.
Scan · a recorded routine, ready to run or publish

02 · Model

Give your business one shared language.

Map each source to the things your business runs on: customers, accounts, orders, claims. People and agents ask in those terms and get the same answer. Swap the CRM? Map the new one once. Everything built on top keeps working.

Studio · DatasetsProduct
A dataset in Studio: a query reads a cases table, and the inspector shows its connection, the UAT environment, and its mapping to the Case business object. A dataset in Studio: a query reads a cases table, and the inspector shows its connection, the UAT environment, and its mapping to the Case business object.

Where your teams work

03 · Experience

Bring agents to the people who do the work.

Ask in plain words. The answer comes back as a live view: the chart, the records behind it, and a drafted next step waiting for your approval. This is the layer your teams will judge us by.

How it works

Follow one request from question to record.

Watch a job run, from the first system it reads to the person who signs off. Add a hiccup. Watch it recover.

Play it through

03·The platform

Studio at the center. Agentic apps where your teams work.

Studio is where systems, data, and agents meet. Agentic apps put that work in front of each team. Scan covers what still lives on screens.

Studio

Where the work gets built.

Connect each system once, model the business, build the agents that use it. Every run goes on the record.

Explore Studio Agentic apps

An experience for every team.

Dashboards you can talk to. Operations that ask before they act. Shaped around one team's work, built with you.

Explore agentic apps Scan

For work on screens.

Show it a task once. An agent repeats it in a real browser.

Explore Scan

Inside Studio

Studio · Run
A dataset run in Studio: it succeeded in UAT in 91 milliseconds, with five recorded steps, their timings, and an event log.A dataset run in Studio: it succeeded in UAT in 91 milliseconds, with five recorded steps, their timings, and an event log.
Every run: steps, timings, events.
Studio · Agents
Creating an agent in Studio: a name, a role, and the runtime it uses in production and in testing.Creating an agent in Studio: a name, a role, and the runtime it uses in production and in testing.
Name an agent, give it a role, pick its runtime.
Studio · Production
A production confirmation: typing PROD confirms a change in the production environment.A production confirmation: typing PROD confirms a change in the production environment.
Production asks you to type its name before anything changes.

04·Industries

The same layers, shaped to the work.

An adviser prepping Thursday's review and an adjuster opening a claim need different screens. Underneath, they share everything.

Live today

Wealth management

Meeting prep, client reviews, and account screening. The advisor signs off on every change.

Explore wealth management

Banking and payments

Every new payee checked. Every release signed.

Insurance

Claims opened from the first notice, coverage drafted for the adjuster.

Commerce and retail

Sellers verified before they go live. Refunds inside your policy.

Asset management

Reconciliation breaks explained, with the fix waiting for approval.

Customer experience

Answers drafted from your knowledge base, each with its source.

Payment approvals

Invoices matched to order and receipt, released on approval.

Identity commerce

Purchases for a verified customer, inside the limit they set.

Your industry, next.

Bring one workflow. We map it with you.

Talk to us

05·On the record

Work that holds up on a bad day.

Systems time out. People step away. Runs survive both, and every step goes on the record with its timing and cost.

Retries
A failed call is tried again after a short wait.
Approvals
A step that needs a person waits for one.
Schedules
A schedule turns on only after a test run passes.
Baselines
Time saved names its baseline, measured or declared.

06·Trust

Built for work that needs a record.

Your risk team will ask about these six first. Each works exactly as written here.

Security in detail
InstanceYour own instance
Own databases, storage, sign-in, and domain.
AccessEvery request checked
Checked against its organization, on every route.
SecretsKeys never shown again
Saved once, never shown or returned.
ApprovalsA person before any change
In Safe mode an agent asks before any step that writes.
RolesAccess by role
Who builds, who runs, who can only read.
AuditA record you can verify
A log sealed entry to entry. Any edit after the fact shows.

07·Getting started

We start inside your team.

Our engineers sit with the people who do the work: connect the first systems, build the first workflow, stay until its value is measured. Then it is yours.

How an engagement works

For developers and agents

Plugs into the systems you already run.

Your systems start runs with a token and read every run back. In preview, AI assistants call your agents as tools over MCP.

Open the developer center
GET /core/api/v1/organizations/{org_id}/runs/{run_id}/events 
# Every step of a run, in order
RUN="$NSTACK_BASE/core/api/v1/organizations/$NSTACK_ORG/runs/$RUN_ID"
curl -s "$RUN/events" \
  -H "Authorization: Bearer $NSTACK_TOKEN"

See it on your own data.

Bring one workflow that matters. We show you all three layers working on it.

Talk to ushello@nstack.ai