THE INTERACTIVE DEMOS

See agents work.
Inside your app.

Give a team of agents a goal, or explore six workflow examples. Make decisions, inspect the shared work, and see how Core fits into your application. No sign-in required.

A PERSISTENT TEAM OF AGENTS

Research to proposal. One team.

Watch the handoffs. Inspect the evidence.
Give the same team a new goal.

NORTHSTAR / PILOT STUDIOGuided simulation · sample data
  1. 01Research
  2. 02Review & revise
  3. 03Pilot proposal
  4. 04Ready for next
OBJECTIVE 01 / RESEARCH

Find a useful pilot for Northstar’s customer onboarding team.

Ready to start
Same team · northstar-pilot3 sessions

Lead

Ready

Researcher

Ready

Writer

Ready

Reviewer

Awaiting approval
Shared context
YOUR BRIEF

Find a useful pilot.

Use Northstar’s sample interviews to turn unclear ownership into an evidence-based opportunity brief.

Optional: see a budget hold and top-up.

Scripted browser simulation with fictional documents. No live model calls or business actions run here.

Swarms implemented · hosted validation pending
What Core handles in this example

Each member has its own persistent session, a pinned agent version and defined permissions. Shared files carry versions; messages name their sender and carry references to the work. Recruitment follows the swarm’s policy, and a request cannot approve itself.

Completing an objective leaves the swarm open and idle. The next objective uses the same members, workspace and sessions. Pause and budget controls preserve work so it can continue.

Swarms are implemented across the runtime, API, SDK and console, with local integration and real-model journeys verified. Hosted replica, recovery and upkeep-job validation remains outstanding. Confirm availability for your preview environment with the team. The research-to-proposal sequence is based on a recorded real-model development journey; this walkthrough uses a shorter, scripted example.

THE PLAYGROUND

Choose a moment.
See what changes.

Start with the experience your users see. The workflow and event trail open as the demo runs, so you can follow every step behind it.

  1. 01Choose a scenario

    Find a moment that fits your app.

  2. 02Make it happen

    Start the run and guide the outcome.

  3. 03Look under the hood

    See the workflow and engineering effort.

Core / Interactive examples SAMPLE DATA · SIMULATED RUN
CATCH ISSUES BEFORE SUBMISSION

Catch the problem before it goes through.

The purchase form is filled in. Run a check to see whether the draft is ready to send.

In your appReady when you are

Purchase request draft

You’re about to request replacement equipment. Check the draft against the sample purchase policy before sending it for approval.

Request
Replacement equipment
Amount
$6,400
See the workflow and events
Preflight check · correction · human confirmation purchase.draft_review_requested
6 connected stepsAGENTS + TOOLS + YOUR SYSTEMS
AgentYour systemWaiting for inputSCROLL TO EXPLORE
Ready when you areChoose a scenario, then press play.0 done
Events to your systems
0 events
Your systems are part of the workflow.

Run a scenario to follow incoming events, live progress, and outgoing notifications.

EVENT PAYLOADAWAITING FIRST EVENT
{
  "event": "workflow.started",
  "run_id": "...",
  "destination": "your-application"
}
ENGINEERING ESTIMATE

What would it take to build this?

Ship one feature in an existing app. Estimated team effort and labor in USD.

Conventional build
10–20 person-hours
$1,183–$2,365
With Core
2.25–4.75 person-hours
$268–$566
First SDK result 10–30 min with a registered tool

Publish an agent with a tool, then call run() whenever you need a result. ensure() publishes the first version and updates it when the definition changes.

import { Jwtson } from '@jwtson/sdk';

// Set JWTSON_API_URL and JWTSON_TOKEN on your server.
const jwtson = new Jwtson();
const agent = await jwtson.agents.ensure({
  name: 'Request reviewer',
  instructions: 'Look up the request and explain what needs review.',
  tools: ['lookup_request'],
});

const { text } = await agent.run('Review request 042.');

The first-result range is a hands-on planning estimate for one server-side agent with one enabled model, scoped credentials, and a registered tool. The feature estimate above also covers app wiring and release checks.

Read the SDK quickstart
Compare the work and assumptions

Same scope for both routes. One purchase form, one evidence rule, one correction step, and confirmed submission to an existing approvals API.

  • Evidence check
    Development
    Conventional build2–4 h

    Implement the quote threshold and amount check as app rules.

    With Core0.5–1 h

    Set the evidence rule and attach the existing quote lookup.

  • Hold, recheck, and submit
    Development
    Conventional build3–6 h

    Persist the draft, resume after correction, and guard against duplicate submission.

    With Core0.25–0.75 h

    Configure a wait, recheck, and confirmation. Bind the existing submission API.

  • App experience
    Development
    Conventional build2–4 h

    Build missing-evidence, confirmation, and receipt states in the form.

    With Core0.5–1 h

    Wire the form to Core progress, input, and confirmation components.

  • Verify the paths
    QA
    Conventional build2–4 h

    Test missing and mismatched quotes, confirmations, and retries.

    With Core0.5–1 h

    Exercise the same policy cases and the app-to-Core submission path.

  • Release and monitor
    DevOps / SRE
    Conventional build0.5–1 h

    Release through the existing pipeline and check failure logs.

    With Core0.25–0.5 h

    Set workflow configuration and verify alerts in the app.

  • Review access
    Security review
    Conventional build0.5–1 h

    Check who can view quotes and confirm requests.

    With Core0.25–0.5 h

    Check the same permissions at the app and tool boundary.

Both routes start with working sign-in, business APIs, data sources, and a release pipeline. Core uses an enabled model where needed. The conventional build coordinates work in app code, with model calls where the experience needs them.

These planning ranges cover initial implementation, release, QA, and security review. Labor uses the rollout calculator’s default rates. Subscriptions, usage, and ongoing maintenance are separate.

Development $125/hDevOps / SRE $125/hQA $85/hSecurity review $150/h

Estimate a wider rollout
Intelligence in the workflow. Results in your world.PARALLEL WORK · BRANCHES · FEEDBACK LOOPS · EVENTS
EXPLORE THE DEMOS

One platform.
Many ways to help.

See how an app can guide someone, check a submission, spot a change, or take on a task.

01
Check a submissionCATCH ISSUES BEFORE SUBMISSION

Catch the problem before it goes through.

Check a purchase request against a sample policy, fix the issue, and decide when to submit it.

Explore this example
02
Market researchTAKE ON THE MANUAL STEPS

One question. Multiple perspectives.

Let specialist agents explore customer needs, market context, and product fit together. Bring the findings back into one actionable recommendation.

Explore this example
03
Customer onboardingGUIDE PEOPLE AS THEY WORK

Meet every customer where they are.

Choose a setup path, coordinate with your own services, and pick up where you left off when the workspace is ready.

Explore this example
04
Campaign studioIMPROVE WORK WITH FEEDBACK

From one idea to every channel.

Bring audience insight and product knowledge together. Refine the creative direction, then turn it into a coordinated set of channel assets.

Explore this example
05
Document studioKEEP PEOPLE IN THE REVIEW LOOP

From your knowledge to a document worth sharing.

Search your knowledge, shape a plan, and watch agents draft and edit. Ask for changes, save the finished document, and tell your systems it is ready.

Explore this example
06
Living knowledgeNOTICE CHANGES AND SUGGEST A NEXT STEP

Notice a change and suggest what to do.

Index model records and approved content once. Help customers find answers in a portal, or give sales operators the same knowledge with a fresh inventory check.

Explore this example
BUILD THE NEXT CHAPTER

Seen the possibilities?
Make them part of your app.

Connect your tools, configure an agent, and build the experience your users need.