Claude Code for enterprise teams · managed preview

Manage the fleet. Understand the spend.

One operating view for work-issued Claude Code runtimes: who is connected, what policy applies, where usage accumulates, what changed, and which actions need attention.

Join the managed preview →See usage controls
Know what is running

One inventory for every managed runtime.

See enrollment, health, uptime, runtime and policy versions, adapter capabilities, group, tags, bounded activity, and observed versus desired state.

AM
ana-macbook-proClaude adapter · online 6h 14m
converged
JK
jon-devboxClaude adapter · policy update pending
drift
RS
rina-laptopClaude adapter · online 2h 08m
converged
TP
teo-workstationClaude adapter · idle 42m
healthy
Manage without screen sharing

Targeted control, from a distance.

Quarantine a managed runtime, refresh policy, synchronize state, collect diagnostics, or request a graceful restart. Commands are signed, expiring, reason-bound, and returned with a bounded result.

↳
refresh_policyEngineering group · issued by Ops
18 / 18
↳
collect_diagnosticsjon-devbox · drift investigation
complete
↳
sync_staterina-laptop · config reconciliation
applied
↳
restart_runtimeana-macbook-pro · graceful lifecycle
signed
Usage & spend accountability

See the usage behind the bill.

Attribute model calls and input/output tokens to the tenant and managed runtime, so cost review can happen by team, project, policy, and operating context.

  • Model-call and token attribution per managed runtime
  • Usage delivery status and failure visibility
  • Policy-aware context for high-consumption activity
  • Context for finance and platform review
Model usage by teamIllustrative · 30 days
Product engineering18.4M tokens
Platform11.2M tokens
Security6.8M tokens
2,481Model calls
31.4MInput tokens
5.0MOutput tokens

Current managed-preview accounting attributes model calls and tokens. Provider-price reconciliation, enforced dollar budgets, and invoice-grade reporting are being completed with design partners.

Engineering policy rolloutRevision 42
01
Review the changeExact diff and bounded simulation
approved
02
Canary two runtimesSigned desired state issued
2 / 2
03
Collect acknowledgementsApplied, partial, or rejected
18 / 18
04
Monitor convergenceDrift remains visible until resolved
healthy
Settings and policy rollout

Change the right settings. Keep a history.

Use authority-signed desired state for policy revision, grouping, tags, runtime target, quarantine posture, and supported Pilot configuration. Safe state changes use optimistic revisions and record who changed what and why.

policy revisiongroup & tagsruntime targetquarantinesafe .pilot state
Control that follows the action

Guardrails for the work, not just the login.

Claude Code becomes manageable when policy reaches the native side effect and the evidence reaches the operator.

01

Native action controls

Map supported Claude tool events into governed process, filesystem, browser, HTTP, trust, message, and file-transfer actions.

02

Approval before consequence

Hold an exact action for expiring human consent, then resume it once with a signed, payload-bound approval.

03

Typed remote recovery

Refresh policy, sync state, collect diagnostics, export receipts, restart, or shut down through signed allowlisted commands.

04

Evidence that joins up

Connect the employee runtime, called resource, policy revision, decision, approval, result, and receipt in one trace.

Adapter boundary

Pilot governs actions surfaced through an installed, supported adapter. It does not provide arbitrary remote shell access, capture unrelated tools automatically, or expose local keys and secrets in the console.

Review architecture →
Built with enterprise teams

Define how Claude Code should run across your company.

Bring your rollout model, provider-billing context, policy requirements, remote-management needs, and evidence obligations.

Design the pilot →Explore the control plane