Teleport is an infrastructure identity platform that gives humans, machines, workloads and AI agents cryptographic identities and uses them to control access to servers, Kubernetes clusters, databases, applications, cloud consoles and MCP servers. Developed by Gravitational Inc. and headquartered in Oakland, California, Teleport began as an open-source project and is led by co-founder and CEO Ev Kontsevoy. The platform replaces static credentials such as passwords, SSH keys and long-lived API tokens with short-lived certificates issued at the moment of access, and records what each identity does. This overview explains how the platform works and what it covers, based on Teleport’s public product documentation and announcements.
At A Glance
| Vendor | Teleport (Gravitational Inc.), headquartered in Oakland, California |
|---|---|
| Category | Infrastructure identity, zero trust access, machine and workload identity |
| Identities covered | Human users, machines, workloads and AI agents |
| Deployment | Teleport Enterprise Cloud (SaaS), self-hosted Teleport Enterprise, or the open-source-based Community Edition |
| Credentials | Short-lived X.509 and SSH certificates; SPIFFE-compatible X.509 and JWT SVIDs for workloads |
| Agentic AI | Secure MCP access, Agentic Identity Framework, Beams trusted runtime for agents |
| Strategic partner | Cisco (strategic partnership and investment announced August 2026) |
What The Platform Does
Teleport organises its platform into four products that share a single identity layer.
Zero Trust Access
Zero Trust Access gives engineers and automation identity-based access to infrastructure through a single access point. Every connection is authenticated with a short-lived certificate tied to the user’s identity, single sign-on and multi-factor authentication, and authorised through role-based access control. Supported resources include:
- Linux and Windows servers over SSH and RDP
- Kubernetes clusters
- Databases including PostgreSQL, MySQL, MongoDB, Redis, Microsoft SQL Server, Snowflake and cloud-hosted database services
- Internal web applications and cloud consoles for AWS, Azure and Google Cloud
- Git repositories and MCP servers
Interactive sessions are recorded and every action is written to a structured audit log. Teleport also supports device trust, so access can be limited to registered, company-managed devices, and hardware security keys for phishing-resistant authentication.
Identity Governance
Identity Governance manages who should have access and for how long. Access Requests let users request elevated roles or specific resources just in time, with approval through Slack, Microsoft Teams, Jira, PagerDuty or the Teleport interface, and access expires automatically. Access Lists group users and grant roles with periodic reviews, and Access Monitoring highlights excessive or unused privileges.
Machine & Workload Identity
Machine & Workload Identity issues short-lived identities to non-human actors such as CI/CD pipelines, automation scripts, services and AI agents. It has two modes:
- Machine ID issues Teleport credentials so that bots can reach resources protected by Teleport, such as running Ansible over SSH or deploying to Kubernetes from a pipeline, without storing static keys.
- Workload Identity issues credentials compatible with the open SPIFFE standard, as X.509 or JWT SVIDs, for workload-to-workload authentication, mutual TLS, and authentication to third-party APIs and cloud providers. Teleport documents federation with cloud services such as Google Cloud Workload Identity Federation.
Identity Security
Identity Security gives security teams visibility of identities and access paths across infrastructure and identity providers. Its Access Graph maps how humans, machines and AI agents can reach critical resources, including access granted outside Teleport, and highlights risks such as standing privileges and unmanaged SSH keys. In July 2026, Teleport announced three additions for agent behaviour: Beams Session Summaries, Agentic Classifiers and Risk Scoring, which classifies SSH, Kubernetes and database sessions by risk level and maps actions to the MITRE ATT&CK framework.
How The Architecture Works
A Teleport deployment is called a cluster and has three main parts.
tsh client
| Component | What it does |
|---|---|
| Auth Service | The Auth Service is the cluster’s certificate authority. It verifies identities, issues short-lived certificates, enforces role-based access policy and stores the audit log and session recordings. |
| Proxy Service | The Proxy Service is the single public entry point for users and machines. It terminates connections, provides the web interface and routes traffic to the right resource over TLS. |
| Teleport agents | Agents run next to protected resources and handle SSH, Kubernetes, database, application and desktop access. They connect outbound to the Proxy Service through reverse tunnels, so protected resources do not need to be exposed to the internet or opened to inbound connections. A Discovery Service can find new cloud resources, such as EC2 instances, EKS clusters and RDS databases, and enrol them automatically. |
Customers can run this architecture as Teleport Enterprise Cloud, where Teleport operates the Auth and Proxy services, or self-host it. Users connect through the web interface, the Teleport Connect desktop app, or the tsh command-line client.
How Machines And Workloads Get Identities
Non-human identities in Teleport are represented as Bots, which are governed by the same roles as people. A lightweight agent called tbot runs alongside each workload, proves the workload’s identity to the cluster, and keeps its certificates renewed.
tbot proves identity through join methods that rely on the platform the workload already runs on, rather than a shared secret. Documented join methods include AWS IAM, Azure, Google Cloud, Kubernetes service accounts, GitHub Actions, GitLab CI, CircleCI, Terraform Cloud and Spacelift, as well as TPM-based joining for on-premises hardware and a bound keypair method for environments without a platform identity.
Workloads receive their credentials either as files written to disk or Kubernetes secrets, or through a local SPIFFE Workload API socket. Workload attestation lets the SPIFFE ID reflect attributes of the calling process, such as a Kubernetes namespace and service account, a Unix user or a systemd service name. On Kubernetes, tbot can be deployed as a daemon set to serve SPIFFE identities to every pod in the cluster.
Agentic AI Capabilities
Teleport treats AI agents as a fourth identity type alongside humans, machines and workloads, and applies the same cryptographic identity, least privilege and audit to them.
Secure MCP Access
Since Teleport 18, Teleport can sit between MCP clients, such as Claude Desktop or Cursor, and MCP servers. Users and agents connect to MCP servers through Teleport with their own identity, role-based access control determines which MCP servers and tools each identity can use, and every tool call is recorded in the audit log. Teleport can also expose protected databases to MCP clients, so an AI assistant can run queries without holding a database password.
Agentic Identity Framework
Announced in January 2026, the Agentic Identity Framework is Teleport’s reference architecture for running AI agents in production. It sets out policies, practices, developer tools and a reference design built on cryptographic identity backed by a hardware root of trust, ephemeral privileges, access guardrails, auditability and real-time enforcement, along with tools to govern MCP access.
Beams
Beams is Teleport’s trusted runtime for agents. It gives each agent an attestable identity and enforces its operational, execution and communication boundaries. Beams Session Summaries condense an agent’s identity, privileges, tool and API calls, prompts, responses and reasoning into a short readable summary. Agentic Classifiers then evaluate agent behaviour against organisation-specific criteria and flag activity inconsistent with the agent’s declared objective.
Availability
Secure MCP access is part of the Teleport platform. Beams Session Summaries, Agentic Classifiers and Risk Scoring were previewed at Black Hat USA 2026 and are available through a technology preview programme, with hands-on customer access scheduled for autumn 2026.
Integrations
| Area | Documented integrations |
|---|---|
| Identity providers | Okta, Microsoft Entra ID, Google Workspace, and other SAML, OIDC and GitHub single sign-on |
| Cloud platforms | AWS, Microsoft Azure, Google Cloud, including automatic discovery of compute, Kubernetes and database resources |
| CI/CD and automation | GitHub Actions, GitLab CI, CircleCI, Jenkins, Terraform Cloud, Spacelift, Ansible |
| Access request workflows | Slack, Microsoft Teams, Jira, PagerDuty, ServiceNow, email |
| Audit and SIEM | Event export to SIEM platforms such as Splunk, Datadog and Elastic |
| AI tools | MCP clients including Claude Desktop and Cursor |
Use Cases And Industries
Teleport publishes use cases for controlling and containing AI agents, vault-free privileged access management, accelerating engineering velocity through faster and more consistent infrastructure access, and identity behaviour and context. Common deployments include replacing shared SSH keys and VPN-based access, giving CI/CD pipelines short-lived credentials, meeting compliance requirements for access reviews and session recording, and issuing SPIFFE identities for service-to-service authentication. In August 2026, Teleport extended infrastructure identity to Linux desktops for developer workstations.