Skip to content
06 · Adoption

Claude: Governance and Cost

Claude governance in two levels, from essential to advanced, with cost under control. Complements your information security and LGPD program.

2 governance levels · cost at every level

Claude Partner Network — Certified Services Partner

Base/labs is Certified in the Claude Partner Network. Our team is Claude certified.

Who it is for

For companies that:

  • already use Claude, or will, and want to start with the minimum governance and move up a level as risk requires;
  • work in regulated industries or have critical AI uses;
  • already have information security or LGPD compliance programs and want to extend them to the use of Claude.
The problem it solves

Without governance, AI use grows without control:

  • overly broad access;
  • data kept longer than needed;
  • spend with no owner;
  • Shadow AI: AI tools used on people's own initiative, with company data, without the company seeing it.

Many companies put governance off because they imagine a project that is too big. Levels solve that: the minimum viable is the starting point, and each level adds controls without redoing the previous one.

Cost has no level: it is there from day one, with an owner, a limit and tracking.

Shadow AI: the AI use the company does not see

Shadow AI is the use of AI tools, including AI embedded in other software, that the company did not choose, approve or even know about. The problems that usually come with it:

  • Company data and personal data sent to tools with no contract and no legal basis (LGPD).
  • Without an inventory, nobody knows which tools and uses exist, and there is no way to know where the risk is.
  • Personal accounts and scattered subscriptions: spend outside the budget and with no owner.
  • Without a policy or an owner, there is no rule on what can be sent to the model.
  • AI outputs used in decisions with no review and no traceability.
  • In an audit, or in front of a client, you cannot prove control over what you cannot see.

The Essential level starts here: an inventory of uses and a one-page policy. Blocking external tools depends on the company's own network and device controls.

Claude plans

Claude plans: what each one already delivers

Team and Enterprise are Claude's own plans, contracted directly with Anthropic or through your cloud marketplace. They are not Base/labs offerings. The table shows what each plan already delivers on its own. Our work, further below, configures and operates those capabilities for your company.

Plan featureClaude TeamClaude Enterprise
Single sign-on (SSO)YesYes
User provisioningAutomatic on first sign-in (JIT)Synced with the directory (SCIM)
Access rolesBasic roles (owner, admin and user)Custom roles and groups
Spend limitsYesYes
Managed settings from the consoleYesYes
Usage dashboard and spend reportYesYes
Usage analytics APINoYes
Audit logsNoYes
Compliance APINoYes
Custom data retentionNoYes
Zero data retention (Claude Code)NoBy agreement with Anthropic, for qualifying accounts; not for chat or Cowork
Organization model restrictionsNoYes
Customer-managed encryption keys and IP restrictionNoYes

For the Essential level, Team or Enterprise will do. The Advanced level's Compliance block assumes Enterprise.

Comparison based on Anthropic's plan and support pages, checked on 11/03/2026. Anthropic may change the plans, so confirm when contracting. Licenses and pricing are handled directly with Anthropic and are not part of our offering. Sources: What is the Enterprise plan? (Anthropic) , What is the Team plan? (Anthropic) , Plans and pricing (Anthropic) .

What we deliver

What we deliver: governance by levels

You do not need to do everything to have the minimum governance. At each level, we deliver the items below, configured and documented for your company. The level is agreed in the proposal, and Advanced includes Essential. Cost applies at every level. Our delivery complements your information security and LGPD program: the focus is Claude, and what the company already has still applies.

Level 1

Essential

Small teams, pilots and low-risk uses. It is the starting point for any company.

We deliver:

For everyone

  • We configure single sign-on (SSO) with the company account and define the process for granting and removing access (automatic with SCIM, at the Advanced level).Team or Enterprise
  • We define the model access route (Claude Enterprise, Console or cloud), that is, where the company will use Claude from.
  • We define the approved MCP connectors: the list of integrations Claude may use to access company systems and data (MCP is the connection standard).
  • We configure the plugins and skills provisioned by the admin, so that they are not installed ad hoc.
  • We build a simple inventory of AI uses and tools in use, including use on people's own initiative (Shadow AI) and AI embedded in other software.
  • We write a one-page usage policy (acceptable use, what must not be sent to the model and care with personal data) and name, with the company, an owner.

If the company uses Claude Code or Cowork

  • We configure the basics of centralized management: sign-in with the company account, the mode that skips permissions turned off, and a deny list for sensitive files and destructive commands.Team or Enterprisealternative on other routes
  • We define the network rules that limit where Claude can connect.Team or Enterprisealternative on other routes
  • We configure execution isolation (sandbox in Claude Code and a virtual machine in Cowork), which limits the folders and network that commands can access.Team or Enterprisealternative on other routes
  • We configure hooks and automated checks for the rules that admit no exceptions.Team or Enterprisealternative on other routes

Level 2

Advanced

Recurring use across several teams, regulated industries and critical uses. The two blocks can be contracted together or separately, and the choice is made in the proposal. The Compliance block assumes the Claude Enterprise plan.

Includes everything from the Essential level

We deliver:

Control at scale

  • We configure the integration of usage and spend with the company's monitoring system, through your own telemetry (OpenTelemetry) or, on Enterprise, through the usage analytics API. It delivers attribution by team and cost center, the data for alerts on unusual use and coverage of cloud usage.Team or Enterprisealternative on other routes
  • We configure automatic user provisioning (SCIM): access is synced with the company directory, and whoever joins or leaves the company gains or loses access to Claude automatically.Enterprise
  • We define roles and permissions by profile and support the client's security review.Enterprise

Compliance

  • We configure the audit trail, with the audit logs and the Enterprise plan's compliance API, where available.Enterprise
  • We configure encryption with customer-managed keys and IP-based access restriction.Enterprise
  • We configure restriction of models and tool versions.Enterprise

Cost, at any level: what we deliver

  • We define the spend baseline and a named budget owner.
  • We configure the native usage and spend dashboard and tiered spend limits.Team or Enterprisealternative on other routes
  • We define the rules for extra usage and overages.Team or Enterprisealternative on other routes
  • We build a simple spend split by area or cost center.
  • For API or cloud usage, we guide the use of prompt caching and batch processing, where applicable.

Badges: Enterprise requires the Claude Enterprise plan; Team or Enterprise requires one of the two plans; alternative on other routes means that, on Console or cloud, it can be done with more work. Items without a badge do not depend on the plan. The badges follow the Team or Enterprise table above.

How we define the level in the proposal

  • The kind of data that goes through Claude: personal and sensitive data call for more control.
  • Industry: regulated industries tend toward Advanced, with the Compliance block.
  • Autonomy of use: chatting is different from using tools and connectors, and from letting agents act in your systems.
  • How many teams use it and whether there are connectors to internal systems: several teams call for the Control at scale block.
  • Audit requirements, from clients, the industry or the regulator.

Moving from Essential to Advanced

We move when Essential is in operation and risk calls for more, and we can start with one of the Advanced blocks. We look at three groups of criteria.

  • Essential maturity: the inventory is up to date, the policy's adherence is tracked, spend limits show no recurring overruns and exceptions are recorded and approved.
  • Operational readiness: there is an owner, support and a team ready to operate the controls of the next level.
  • Risk infrastructure: security, data protection and operational controls of the current level are working and verified.
You leave with
  • The agreed governance level, documented and applied.
  • Spend control with defined owners, at any level.
  • A clear path to move up a level, without redoing what exists.
How we measure

We track spend per tier against the limit, approved exceptions and adherence to the policy of the agreed level.

What we need from you
  • Security, identity and privacy teams taking part in the decisions for the agreed level.
  • A budget owner.
  • Named admins to operate the policy.
FAQ

Frequently asked questions

Does this replace our information security or LGPD program?

No. The offering complements what the company already has. The focus is Claude: we configure the product's controls and align its use with existing policies and processes, such as those for information security and LGPD compliance. General LGPD compliance and the legal framing stay with your legal and privacy team.

Do I need to do everything?

No. The Essential level already gives the minimum governance, and you move to Advanced, whole or by block, only when risk calls for it. The cost part applies at every level.

How do I choose the level?

The level is agreed in the proposal, before we start. In the scoping conversation, we look at the kind of data that goes through Claude, the industry, the autonomy of use (chatting, using tools and connectors, or letting agents act in your systems), how many teams use it and whether there is an audit requirement. Small teams and low-risk uses usually stay at Essential. Regulated industries choose Advanced with the Compliance block, and companies with several teams but no regulatory requirement can choose only the Control at scale block.

Can I move up a level later?

Yes. Advanced includes Essential and adds controls, without redoing what is already configured, and you can contract one block at a time.

Do the levels depend on the plan we contract?

Some controls depend on the plan and the model access route. For example, console-delivered managed settings and spend limits per organization, group or member require Team or Enterprise; in the Console, limits are per workspace; on a cloud, they live with the provider. We define the level that is possible together with the access route.

Does Anthropic define governance levels?

Not that we found. Claude Code's administration documentation organizes controls as decisions, not as levels, and Anthropic's enterprise guides treat governance as part of the initial phase. The levels in this offering are Base/labs's own and apply the idea of risk-proportionate controls, as in the NIST AI RMF.

What about LGPD and AI regulation in Brazil?

LGPD already applies to the use of personal data with AI: a purpose compatible with collection, informing data subjects and care with sending data to tools abroad. The AI legal framework, bill PL 2,338/2023, is still going through the Chamber of Deputies and, according to the versions under debate, provides obligations by use-case risk level. We follow the text, and the levels were designed to adapt to it. This is not legal advice: the framing is defined with your legal team.

What about the AI use people already do on their own (Shadow AI)?

It is the starting point of the Essential level. We begin with an inventory of AI uses and tools in use and define a short policy on what can and cannot be sent. The inventory shows what exists, and the approved access route gives an alternative to tools used on people's own initiative. Blocking external tools depends on the company's network and device controls, which stay with your security team.

Is it suitable for regulated industries?

Yes, with the Advanced level and the Compliance block: an audit trail, encryption with customer keys, restriction of models and IP-based access and alignment with your LGPD policies.

Related offerings

Want to talk about this offering?

We start with a conversation to understand your scope. Each offering has a closed scope and price per phase.