Transition Plan

Both consoles run in parallel on the same data. Governance moves one enforcement target at a time, and each move is reviewed, reversible and proven before it enforces.

Ground rules

• Observe pages are safe from day one. Overview, Costs & Savings, Findings & Interactions, Users, Inventory, Agents and Dashboards read the same data V1 reads. Use them now; nothing in V1 changes.
• Data is reflected in both places. Applications, keys, servers, findings and interactions created or observed in either console appear in both. There is no import and no sync step.
• Governance moves per target, not per console. A target is governed either by its legacy settings or by the Policy Engine, never both. Switching one target changes nothing for the others.
• Every switch starts with a review and ends with a snapshot. Conversion is previewed before anything changes, activation snapshots the legacy configuration, and disabling restores it.

Phase 1: the gateways

Start here. The LLM Gateway and MCP Gateway targets are the most mature: full control surfaces, conversion review, simulation, replay and rollback.

1. Tidy in V1 first. Resolve duplicate application names and disable providers you no longer use.
2. Set model prices. Under Settings, Models, confirm input and output prices for every model you route to.
3. Review the conversion, do not activate. Read the generated rules and the scope-by-scope changes.
4. Activate and change nothing. Revision 1 is your current behaviour. Let it run for a few days.
5. Repeat for MCP Gateway. Tools, Guardrails, Token saving and Group & User Rules from each server arrive as sentences in the ten MCP cards.
6. Add new controls on monitor, then replay before you enforce.

Full detail is in Switching from settings.

Phase 2: the device sensors

Endpoint Agent and Browser Extension control sets are still undergoing minor changes. Their data is already in V2: device status under Users, agentic inventory under Agents and Inventory, interactions in the feed. Keep authoring their controls where you do today, and do not switch the Endpoint Agent target yet. Quilr will confirm when they are settled.

Phase 3: data, access and the last of V1

1. Publish retention before enforcement starts. Retention settings can be published and reviewed today, but enforcement is not yet active and nothing is purged. Once enforcement is on, a published horizon applies to data already collected.
2. Move exports and alerts. Recreate recurring extracts as Export Center definitions and check that Slack, SIEM and webhook integrations show as installed.
3. Optionally migrate access control. Roles & Permissions offers a wizard from V1 authorization to V2 bindings. V2 is default-deny once enabled, so run the wizard's review first.
4. Retire V1 habits, not V1 data. V1 remains available and reads the same data. Nothing needs exporting or copying.

Where to do what during the transition

Task	Before Phase 1	After Phase 1
Look at activity, findings, users, assets, cost	Either console, V2 recommended	Either console, V2 recommended
Create LLM apps, issue keys, add providers	Either console	Either console
Register MCP servers, connections, OneMCP	Either console	Either console
LLM guardrails, limits, routing, identity, token saving	V1 or V2 settings	Govern, Policy Engine, LLM Gateway only
MCP tools, guardrails, token saving, group and user rules	V1 or V2 settings	Govern, Policy Engine, MCP Gateway only
Endpoint Agent and Browser Extension controls	Where you do today	Where you do today, until Phase 2
Detection models	Either console	Either console, builder in V2
Retention, exports, dashboards, skills, models catalog	V2 only	V2 only

The one-way step, restated

Activating a target is reversible: disabling restores the frozen legacy snapshot exactly as it was. What is not reversible is the work you do under the engine afterwards. It stays in revision history and does not reappear on the legacy settings screens. Treat disable as a rollback to the moment you switched, and use the engine's own rollback for everything after.