Connections for Frontier Investigation Agents
Connect your stack. Give every agent the full picture.
Secure, read-only connections bring signals from your clouds, observability, code, databases and collaboration tools into one context, so Frontier Investigation Agents can find the why before anyone acts.
Production access, without lowering the bar.
Connections meets the security floor your team has already approved. It never asks you to lower it for an AI tool.
100+ connections live today. Any MCP server, too.
Clouds, data, observability, DevOps and incident tools in one context, so agents see the whole system at once.
Clouds, data and Kubernetes
AWS, Google Cloud, Azure, Vercel and Kubernetes. PostgreSQL, MySQL, MongoDB, Redis, Kafka, Elasticsearch and more.
Observability and incidents
Datadog, Grafana, Prometheus, New Relic, Dynatrace and more. PagerDuty, ServiceNow and Atlassian for incidents and tickets.
Anything with an MCP server
Tools that expose an MCP server connect through the Model Context Protocol, so new connections land in minutes. Wiz, Snyk, Kubecost, Terraform and more are on the roadmap.
Enterprise-grade security at every layer.
Pick the network tier your security team already trusts. Connections brokers identity and access for every task.
Four network tiers
Public HTTPS, IP allowlist, AWS PrivateLink VPC endpoint or site-to-site VPN, picked per workload.
Scoped credentials at task time
A short-lived, scoped token lives in the sandbox, not in the prompt or a static config, and is revoked when the task completes. TLS in transit, AES-256 at rest.
An approval gate on every action
In Manual, a human approves each action. In Auto, pre-approved runbooks run. Full audit logs, RBAC and SOC 2 Type II.
Questions, answered.
What is CloudThinker Connections?
CloudThinker Connections is the secure production-connection layer that lets AI agents reach your infrastructure without lowering your security posture. It supports four network tiers (public HTTPS, IP allowlist, AWS PrivateLink VPC endpoint and site-to-site VPN), each with brokered identity, scoped credentials issued at task time, and tamper-evident audit.
What network tiers does Connections support?
Four tiers, picked per workload: public HTTPS (cloud APIs, SaaS), IP allowlist (IP-gated databases), AWS PrivateLink VPC endpoint ("no public internet" workloads), and site-to-site VPN (on-prem and hybrid). You pick the tier that matches the security floor your security team has already approved. Connections does not ask you to lower it for an AI tool.
Does CloudThinker hold long-lived production credentials?
No. Connections brokers a short-lived, scoped token at task time. The credential lives in the execution environment (the sandbox), not in the prompt and not in a static config. Once the task completes, the credential is revoked. There is no .env on a developer laptop with admin access to production.
Which connections does CloudThinker support?
100+ native connections live today across six areas: cloud and platforms (AWS, Google Cloud, Azure, Vercel), databases and streaming (PostgreSQL, MySQL, MongoDB, Redis, Neon, Apache Kafka, RabbitMQ, Elasticsearch), observability (Datadog, Grafana, Prometheus, New Relic, Dynatrace, AppDynamics, Zabbix, Better Stack, Coralogix, SigNoz, Rollbar), infrastructure and DevOps (Kubernetes, Cloudflare, Argo CD, Jenkins, CircleCI, AWX, SonarQube, Vault, SSH), incident and ticketing (PagerDuty, ServiceNow, Atlassian, Backlog), and IoT (Flespi). On the roadmap: observability (Sentry, Honeycomb, Chronosphere, Sumo Logic, Splunk, Moogsoft), infrastructure and IaC (Terraform, Ansible, Spacelift, env0, Firefly, Backstage), security and DevSecOps (Wiz, Orca Security, Prisma Cloud, Lacework, Aqua Security, Sysdig, Snyk, GitGuardian, CrowdStrike, Keycloak, Firebase), cost and FinOps (Kubecost, CloudHealth, Cloudability, Vantage, Spot.io, ProsperOps, nOps, CAST AI), and more incident tooling (incident.io, Rootly, FireHydrant, Opsgenie, BigPanda). Beyond the native catalog, any tool that exposes an MCP server connects through the Model Context Protocol, so net new connections land in minutes rather than weeks of platform engineering.
Explore the platform.
- Platform OverviewOne unified layer under every agent.
- AI-DLCAI writes the code. Agents run it in production.
- Auto ModeSafe autonomy for cloud operations, on your terms.
- Self-hosted sandboxesCloudThinker, on your own machine.
- Skills FrameworkTurn operational knowledge into safe agent action.
- Security & ComplianceSecurity by design, at every layer.
Ready to connect your stack?
Connect in minutes. Frontier Investigation Agents then find the why across your whole stack.