Frontier Governancefor AI Agents

Guardrails, scoped identities and audit trails for every agent you run.

Every agent in your company works through connections you approve, under written rules, with a record of what it did. Secrets stay encrypted, and with Vault connected, agents issue short-lived credentials through your approvals. Risky writes wait for a person, and every action lands in a read-only audit log your auditors can check.

  • Cloud AI agents
  • Coding agents
  • MCP servers
  • Identity
  • Secrets
  • ITSM, CMDB
  • SIEM
  • Team chat
  • CloudThinker Resolve
  • Policies
  • Approvals
Existing systemsNew with CloudThinkerRead-only connection

The work behind every agent

Teams are adding agents every month.Security still can’t say what they touch.

01The manual work

Review

Security reviews each new agent by hand, hands out long-lived keys and pieces together logs after something goes wrong.

02The agent handoff

Governed by default

Frontier agents work through scoped connections, encrypted credentials and written rules, with every action recorded as it happens.

03Your engineers’ role

Set policy

Decide what each agent may do alone, what needs approval and who approves. Change it as trust grows.

Where CloudThinker fits

Your stack stays.Agents work inside it.

01Agents in use

Already running

  • AWSAmazon Bedrock agentsModels and agents on AWS
  • AzureAzure AI Foundry agentsAgents built on Azure
  • Google CloudVertex AI Agent BuilderAgents built on Google Cloud
  • GitHubCoding agentsAgents that open pull requests
  • MCPMCP serversTools your agents can call

No change to your agents

02Identity and secrets

Your system of record

  • Okta and Microsoft Entra IDWho owns each agent
  • VaultHashiCorp VaultSecrets and dynamic credentials
  • AWSAWS IAM Identity CenterRoles and permission sets
  • ServiceNowChange and approval records

Source of truth stays put

03Governance

CloudThinkerCloudThinker

  • ScopeOnly the connections you approve
  • IssueShort-lived credentials from Vault
  • GateRisky steps held for a person
  • RecordEvery action in a read-only audit log

Least privilege by default

04Oversight

Approvals, not surprises

  • SlackSlackApprove or narrow a held action
  • Microsoft TeamsMicrosoft TeamsApprovals inside the thread
  • JiraPolicy exceptions tracked as tickets
  • SplunkAudit log exported as CSV

Every action attributable

Logos show common stacks. CloudThinker governs each agent through the identities and approvals you already run.

Example scenario

Wednesday, 15:20. An agent asks for production.A rule holds it for a person.

A 150-engineer logistics platform running 6 kinds of agents across AWS and EKS, with Okta, Vault and a CISO who has to sign off on every new one.

  1. 15:20

    Agent requests an actionAgent

    During an incident, an agent proposes scaling payments-api to 0 replicas in prod-eu to stop a retry storm.

  2. 15:20

    Policy check runsAgent

    An Ask first rule covers scaling in production, so the write is held for a person. Nothing runs yet.

  3. 15:21

    Context sent to the approverAgent

    The on-call approver gets the action, the evidence behind it, the blast radius and the rollback command.

  4. 15:24

    Approver narrows itYour team

    Rejects scale-to-zero, approves scaling to 2 replicas and pausing the retry queue instead.

  5. 15:24

    Short-lived credential, one changeAgent

    The agent issues a short-lived credential from Vault under the approval and runs only the approved change.

  6. 15:31

    Recorded for the CISOAgent

    The request, the approval decision, the action and its outcome are in the audit log, with the agent named as the actor.

#agent-approvals4 messages
  • CloudThinker15:21

    Held for approval: scale payments-api to 0 in prod-eu. Reason: retry storm from queue settlement-retry, 18k msgs/min. Blast radius: EU checkout. Rollback: scale to 6.

  • On-call approver15:24

    Not zero. Scale to 2 and pause settlement-retry.

  • CloudThinker15:26

    Done with a short-lived Vault credential. payments-api is at 2 replicas and settlement-retry is paused. Error rate back to 0.3%.

  • CloudThinker15:31

    Audit log updated: request, approval, action and outcome, with this agent as the actor.

from request to a narrower, approved action
4 min
writes run without a rule check
0
audit trail with the full story
1

An illustrative example. Team, systems and times are representative, not a specific customer.

Frontier governance

Every agent has an identity.Every action has a record.

One control plane for every agent you run, so a new agent arrives with the same connections, rules and audit trail as the last one, and autonomy grows one approved step at a time.

Scope every connection
Agents reach only the connections you approve, with the access you grant each one, such as a read-only cloud role.
Protect every credential
Secrets are stored encrypted and passed in only when a command runs. With Vault connected, agents issue short-lived credentials through your approvals.
Gate risky actions
Written rules decide what runs on its own, what waits for approval and what is refused, per connection.
Record everything
Changes, approval decisions and command results written to a read-only audit log, exported as CSV for your auditors.

What changes

Same team. Same tools.Far less of the work by hand.

A team adds an agent
TodayA one-off security reviewWith frontier agentsOnboarded on the shared policy
Agent access
TodayLong-lived keys with broad scopeWith frontier agentsScoped connections, every write checked by rules
Risky actions
TodayAllowed or blocked by guessworkWith frontier agentsDecided by written policy, with approvals
After an incident
TodayLogs pieced together by handWith frontier agentsA read-only record of every step
Audit
TodayScreenshots and spreadsheetsWith frontier agentsEvidence exported from the log

Integrations

Connects to the rest of your stack.Read-only to start.

  • AWS IAM
  • AWS CloudTrail
  • Amazon Bedrock
  • Amazon EKS
  • Okta
  • AWS IAM Identity Center
  • GitHub
  • GitLab
  • Slack
  • Microsoft Teams
  • ServiceNow
  • Jira
  • Splunk
  • Datadog

Adoption path

One pilot.Then company-wide.

The rollout follows the four phases of the AWS Cloud Adoption Framework, so it fits the plan your cloud team already runs.

  1. 01Envision

    Map the agents you run

    List every agent, what it can reach and how it authenticates. Start with the one closest to production.

  2. 02Align

    Write the policy

    Agree what agents may do on their own, what needs approval and who approves, per connection.

  3. 03Launch

    Move agents onto the control plane

    Move agent secrets into encrypted credentials and Vault, and run every agent on the same rules and audit trail.

  4. 04Scale

    Make it the default

    New agents can only reach production through the control plane. Autonomy grows as each agent earns it.

Customer proof

MSM.Results on the record.

A Vietnamese digital transformation partner moved from AI-assisted delivery to governed AI agents, with approvals on risky changes and an audit log of every agent action.

Read the case study
lower tooling cost through consolidation
63%
for manual tasks that used to take hours
Minutes

Trust and control

Agents do the work.Your team keeps control.

Approval by policy
Agents propose. You decide what runs on its own and what waits for approval, and a short list of destructive calls never runs.
On the record
Every check and action stays in the investigation, with who approved it. Org-wide audit logs come with Scale and Enterprise.
Independently audited
SOC 2 Type II report available in our Trust Center. ISO/IEC 42001 audit in progress.Visit the Trust Center
Runs where you need it
SaaS today, also through AWS Marketplace. Dedicated and self-hosted deployments are in early access.

Questions

What teams askbefore they start.

Can we start without giving agents write access?
Yes. Most teams start read-only, with every write action held for approval. Autonomy grows one action type at a time, as the record shows it is safe.
Where do credentials live?
Connections hold the access you grant, such as a read-only IAM role. Other secrets are stored as encrypted workspace credentials and passed in only when a command runs. With Vault connected, agents can issue short-lived database or cloud credentials through your approvals.
Who decides what an agent may do?
Your team does, in written rules per connection. Reads always run. Each write runs on its own, waits for an approver, or is refused.
Can auditors check the log?
Yes. The audit log records changes, approval decisions and command results, and nobody can edit or delete an entry, Owners included. It keeps 90 days of events, so export it as CSV on a schedule for SOC 2 or ISO reviews. Audit logs come with Scale and Enterprise.

Give every agent an identity.Keep every action on the record.

Start with the agent closest to production. See its access, policy and audit trail in one place before you scale to the rest.

  • A CloudThinker team member holding a card reading "up to $200K active AWS credits"

    Up to $200K in AWS credits

    Applied to your own AWS account.

  • A CloudThinker team member presenting the AWS Partner AI Services Competency badge for Agentic AI Consulting Services

    AWS AI Services Competency

    Validated for Agentic AI Consulting.

  • An engineer approving a request beside a global operations map, an uptime dial, and GDPR and SOC compliance marks

    Covered 24/7, inside your policy

    With a SOC 2 Type II report.