# Start from something that already works.

**Start from something that already works.**

Template-first, never a blank canvas. Describe your outcome and KrewOS recommends a Crew Template matched to your industry, jurisdiction, language, risk level and required outputs — then shows you exactly what it would change.

Full Studio, Hub and Evaluation Lab on the trial. Development environment only.

KrewOS Studio

[Start free](/signup) [Watch a run](/demo)

## The Template Wizard picks the starting point.

Recommendations are made against declared criteria, not vibes. The wizard shows why each candidate was ranked where it was, so a domain expert can overrule it with a reason.

| Criterion | What it selects on | Example |
| - | - | - |
| **Industry** | Classification tags on the template and its Pack | Legal services → LawCrew Document Review |
| **Jurisdiction** | Declared jurisdictional coverage and currency date | Singapore → SG Legal Practice Pack templates first |
| **Language** | Supported input and output languages | English + Simplified Chinese → Content Studio bilingual variant |
| **Risk level** | Human-control level the template enforces | Regulated → only templates with a permanent full-review floor |
| **Required outputs** | Typed output contract of the crew | Approved workbook → Finance Close |
| **Connectors available** | Declared connector dependencies against your registry | No CRM connected → BizDev Intelligence flagged, not hidden |

## Assemble Mode: the editor refuses bad structure.

Typed ports will not connect if the contracts do not match. Nested Pods collapse to one node so a twelve-agent crew stays readable. Risk, external side effects, missing permissions and estimated cost are visible on the node, not buried in a settings panel.

**Typed ports**

Invalid connections are rejected at draw time with the contract mismatch named.

**Collapsible Pods**

A Pod is one node until you open it. Complexity stays in the manifest, not on screen.

**Effect flags**

Nodes that cause an external side effect are marked, and marked again if no approval gate precedes them.

**Cost estimate**

Per-node credit estimates roll up to a per-run figure before you have run anything.

### Describe the change. Review the diff. Then decide.

Plain-English changes never edit production. They produce a change plan, a visual diff, a validation result and a test run. You accept all, accept some, revise or reject.

```
You: add a second reviewer for matters over S$2m
      and require both to approve before release

Change plan — 3 changes, 0 breaking
+ node  approval.senior_partner  (human gate)
~ edit  approval.reviewer.policy
          quorum: 1 -> 2 when matter_value > 2000000
~ edit  release.gate.requires
          [reviewer] -> [reviewer, senior_partner]

Validation   12 checks passed, 0 failed
Test run     sample case SG-2291 · 2 gates hit · 41 credits
Effects      unchanged — no new external actions

  [ Accept all ]  [ Accept selected ]  [ Revise ]  [ Reject ]
```

Nothing in a change plan touches a deployed release. Applying a plan produces a new draft version that still has to pass the evaluation gate to be promoted.

**Engineer Mode**

The same crew as manifest, schemas, prompts, policies, Git status and generated project files. Edit here and the canvas follows.

## Test Mode, before anyone else sees it.

**Sample inputs**

Run against packaged sample cases or your own, step through node by node, and inspect the evidence each node retrieved.

**Version comparison**

Run two versions over the same case set and compare outputs, evidence sufficiency, cost and gate outcomes side by side.

**Evaluation suites**

Attach a suite and set the release gate threshold. Promotion is blocked until the suite passes at the declared bound.

**Guardrails built into the editor**

These are caught before you can test, not after you deploy. Each one names the offending node and offers the standard remedy.

Unbounded loops Missing exit paths Orphan nodes Excessive permissions Side effect without an approval gate Untyped port Undeclared external write No evidence floor set Knowledge pack past its freshness window

---

Canonical page: https://dev.krewos.ai/studio