Teleport and Formal both control SSH, Kubernetes and database access. Formal's Connector is a protocol-aware reverse proxy in your VPC that masks query results inline and applies the same Rego policies to engineers and AI agents.
Formal vs Teleport: the Teleport alternative at a glance
Criteria
Formal
Teleport
Access
Architecture
✓Documented: Reverse proxy in your VPC reads each request
✓Documented: Agents check the roles in each user's certificate[2]
Details: Architecture
Formal
The Connector, a reverse proxy container in your infrastructure, sits between identities and resources; the Endpoint runs on laptops and agent hosts. The Control Plane, hosted by Formal, cannot interact with your data.
Teleport
The Auth Service and Proxy Service form the control plane. Teleport Agents in your infrastructure proxy traffic and check the roles encoded in each user's certificate.[2]
Postgres, MySQL, MariaDB, Redshift, MongoDB, Snowflake, BigQuery, ClickHouse, DynamoDB, Redis, S3, AWS and GCP APIs, HTTP, gRPC, web apps, MCP servers, LLM APIs, SSH and AWS SSM, RDP, Kubernetes. Formal proxies supported native database protocols, including PostgreSQL, MySQL, MongoDB, and Snowflake, so agents can use database clients without wrapping each operation as a tool.
Teleport
MCP servers, applications, databases, Kubernetes clusters, Linux and Windows servers, Windows desktops, AWS, Azure and GCP consoles and resources, GitHub. Teleport supports native clients for enrolled databases, with identity-based access, database roles, and query auditing for supported engines.[3]
Upstream credentials
✓Documented: The Connector holds them; users never do
✓Documented: Short-lived certificates from its own CA[2]
Details: Upstream credentials
Formal
The Connector authenticates upstream with a Native User (password, SSH key, cloud IAM identity or kubeconfig). Users sign in with their Formal identity and never need it. Formal brokers configured database credentials, model-provider keys, API tokens, and MCP OAuth tokens at the proxy, keeping those upstream secrets out of the agent process.
Teleport
The Auth Service maintains certificate authorities that sign short-lived host and client certificates. Teleport replaces many static client credentials with short-lived certificates and brokers supported upstream access. Clients hold the Teleport-issued identity needed to connect.[2]
Just-in-time access
✓Documented: Slack approval, exception that expires
◐Partial or edition-limited: Identity Governance; Community: CLI only[3]
Details: Just-in-time access
Formal
Slack forms and Workflows grant time-bound policy suspensions. Employees can request a one-time or time-bound exception (30 minutes to 7 days) on a blocked action. Agents and operators can request time-bound resource access through approval workflows. Protocol policies continue to apply during supported sessions after access is granted.
Teleport
Access requests and reviews in Teleport Identity Governance (Slack, PagerDuty, Teams, Jira, ServiceNow). Community Edition requests roles through the CLI only. Teleport Access Requests provide temporary role or resource elevation with configurable approval and expiration, integrated with short-lived certificate-based identity.[3]
Device and workload context
✓Documented: Identity, device and workload context
✓Documented: Device trust and workload attestation[18]
Details: Device and workload context
Formal
Policies can use identity-provider attributes, Endpoint device information, and supported agent process or workload identity context to decide what access is appropriate.
Teleport
Teleport supports device trust and identity-based roles. Machine and workload identity can attest supported environments before issuing short-lived credentials.[18]
Private access for agent workloads
✓Documented: Identity-based private access
✓Documented: Access to enrolled infrastructure[18]
Details: Private access for agent workloads
Formal
Endpoint and customer-hosted Connectors provide identity-based access to protected private resources, with supported protocol policies applied within the connection.
Teleport
Teleport provides authenticated access to enrolled infrastructure, including SSH, Kubernetes, databases, desktops, and applications, through its proxy and resource agents.[18]
Inside the session
Controls inside a database session
✓Documented: Block or rewrite each query before it runs
◐Partial or edition-limited: Role and table permissions; PostgreSQL only for tables[5]
Details: Controls inside a database session
Formal
Rego policies run at session and request stages: block or rewrite a query before it reaches the database. Formal evaluates policies on supported database statements. An agent can be allowed to SELECT while destructive operations such as DROP or DELETE are blocked.
Teleport
Role-based access by database labels, database users and database names, plus table, view and procedure permissions for PostgreSQL. Teleport provides database object access controls and database-native permissions, with requirements varying by engine and version. This differs from general SQL request transformation.[5]
Response-stage policies mask columns (nullify, hash, partial redaction, fake data), filter rows or block the response before it reaches the client. Formal can mask sensitive database fields and filter returned data using identity and policy, reducing the sensitive information available to an agent and its model context.
Teleport
Masking query results: not documented in Teleport's database access docs. Teleport constrains database access through roles and object permissions. Inline result masking, policy-driven row filtering, and SQL rewriting are not documented equivalents.[5]
Session recording
✓Documented: SSH and kubectl exec recorded, risk-scored
✓Documented: Playback and live view; moderation in Enterprise[3]
Details: Session recording
Formal
Every SSH session and every kubectl exec and attach session is recorded with a risk score, and every query and command is logged. Recordings open in the Sessions app. Formal provides SSH and Kubernetes recordings with playback, plus database statement, remote MCP, and LLM request logs attributed to users and agents.
Teleport
Recording with playback and live session view for SSH, Kubernetes, desktops and databases; session sharing and moderation in Enterprise editions. Teleport records supported SSH, Kubernetes, desktop, and database sessions. Beams adds inference audit, and recordings can use customer-controlled storage.[3]
SSH and Kubernetes actions
✓Documented: Protocol-specific action policies
✓Documented: SSH and Kubernetes roles and recordings[16]
Details: SSH and Kubernetes actions
Formal
Formal proxies supported SSH and Kubernetes sessions and evaluates protocol-specific commands or requests, with identity attribution and activity records for agent operations.
Teleport
Teleport provides SSH and Kubernetes access controls, short-lived credentials, and supported session recordings, including access used by agent workloads.[16]
HTTP and API policies
✓Documented: HTTP request policies and rewriting
◐Partial or edition-limited: Enrolled HTTP applications[18]
Details: HTTP and API policies
Formal
Formal applies policies to supported HTTP requests, including host and operation restrictions, header or body rewriting, and configured credential injection for APIs agents call.
Teleport
Application Access protects enrolled HTTP applications and services. General rewriting and secret injection for arbitrary outbound agent API traffic are separate requirements.[18]
Request rewriting
✓Documented: Protocol-specific transformations
◐Partial or edition-limited: Shared SQL and LLM rewriting not documented[15]
Details: Request rewriting
Formal
Formal supports protocol-specific transformations, including SQL rewriting, HTTP header changes, and LLM request-body rewriting, through the same policy language.
Teleport
Teleport provides protocol-specific access controls and an inference proxy. Arbitrary SQL and LLM message-body transformation through one policy engine is not documented.[15]
Inline action enforcement
✓Documented: Inline protocol policy enforcement
✓Documented: Infrastructure, database and MCP policies[14]
Details: Inline action enforcement
Formal
Formal evaluates policies inline on supported database, remote MCP, HTTP, and model traffic. Policies can deny operations, restrict tool calls, and mask or block sensitive content independently of Insights findings.
Teleport
Teleport enforces supported infrastructure roles, database object permissions, and MCP tool allow/deny policies on traffic routed through its proxies.[14]
AI agents
AI agents
✓Documented: Agent identity tied to its human; MCP gateway
◐Partial or edition-limited: MCP proxying; Enterprise owner attribution[3]
Details: AI agents
Formal
The Endpoint trades OIDC tokens for short-lived Formal credentials, and policies see the human behind a machine session. The MCP gateway adds the upstream credential and logs every tool call.
Teleport
MCP server proxying, identity-based access and per-tool audit events in every edition. Per-agent certificates and agent-to-owner linkage in Enterprise editions. Teleport supports machine and workload identities. Beams adds delegated agent identities with scoped permissions tied to a human or another agent.[3]
Direct traffic outside MCP
✓Documented: Supported direct traffic inspection
◐Partial or edition-limited: Enrolled resources and Beams paths[15]
Details: Direct traffic outside MCP
Formal
Endpoint can inspect supported model, API, database, and remote MCP traffic selected by network rules, including calls from subprocesses outside a configured MCP client.
Teleport
Teleport governs enrolled resources and Beams traffic routed through its proxies. It does not automatically inspect arbitrary external API traffic from every developer process.[15]
Formal routes supported remote HTTPS MCP connections through its proxies, with server access policies, configured upstream authentication, and tool-call logging.
Teleport
Teleport enrolls MCP servers as protected resources and brokers authenticated access through its proxy, with MCP-aware permissions and audit.[14]
Formal evaluates policies on supported remote MCP tool calls and context. An agent can be allowed to read an issue while a write or merge operation is blocked.
Teleport
Teleport roles allow or deny individual MCP tools using names, patterns, and identity traits. These controls apply to calls routed through enrolled servers.[14]
LLM traffic inspection
✓Documented: Model request and response inspection
✓Documented: Beams inference inspection and audit[15]
Details: LLM traffic inspection
Formal
Formal inspects supported model requests and responses routed through its proxies and can apply policies using provider, model, request, response, and token-usage context.
Teleport
Beams LLM Proxy inspects and audits inference requests and responses routed through it. Coverage is associated with the Beams runtime and configured inference path.[15]
Prompt and response masking
✓Documented: Mask or block sensitive content
◐Partial or edition-limited: Inference audit; validate content redaction[15]
Details: Prompt and response masking
Formal
Formal can classify sensitive content and apply masking or blocking policies to supported LLM and remote MCP messages before protected data reaches the next service.
Teleport
Beams provides request and response inspection and audit. Validate content-redaction requirements separately; native SQL result masking is not equivalent to LLM auditing.[15]
Prompt-injection detection
✓Documented: Log-based Insights analysis
◐Partial or edition-limited: Audit and containment, not classification[15]
Details: Prompt-injection detection
Formal
Formal Insights analyzes recorded LLM traffic for prompt injection and other AI risks, producing findings linked to the relevant activity. This is log-based analysis rather than inline prompt classification.
Teleport
Beams provides inference audit and runtime containment. A native prompt-injection classifier is not established by those access and logging controls.[15]
Shadow AI discovery
✓Documented: Discovery from supported proxied traffic
◐Partial or edition-limited: Enrolled resource and runtime visibility[15]
Details: Shadow AI discovery
Formal
Formal discovers recognizable agents, remote MCP services, and skill-related activity from supported proxied traffic on managed devices and sandboxes, including unregistered services.
Teleport
Teleport supports discovery and enrollment of resources and MCP services. Activity outside enrolled resources and configured agent runtimes requires another observation path.[15]
Operations
Deployment
✓Documented: ECS Fargate or Helm, in your cloud
✓Documented: Teleport-managed cloud or self-hosted[3]
Details: Deployment
Formal
Linux, AMD64 or ARM64. Recommended on AWS ECS Fargate or Kubernetes with Helm (EKS, GKE, AKS, on-premises). AWS PrivateLink to the Control Plane is available. Formal Endpoint runs on managed machines or sandboxes and Connector runs in the customer environment. Routing, client TLS trust, and endpoint enforcement determine network coverage.
Teleport
Teleport-managed cloud for Enterprise (Cloud); self-hosted for Enterprise (Self-Hosted) and Community Edition. Teleport offers self-hosted and managed infrastructure-access deployments. Beams runs in its own managed runtime; assess that deployment separately from a Teleport cluster.[3]
Policy management
✓Documented: Shared Rego policy language
✓Documented: Roles, labels and MCP configuration[14]
Details: Policy management
Formal
Formal uses Rego across supported protocols, with protocol-specific inputs and actions. Policies can be managed through Git and infrastructure tooling and backtested against recorded traffic.
Teleport
Teleport roles, access requests, resource labels, and MCP permissions are managed through configuration, APIs, and infrastructure tooling.[14]
✓ Documented◐ Partial or edition-limited✕ Not documented
Formal publishes this comparison. Every Teleport fact links to Teleport’s own pages, listed under Sources with the date we read them. Updated Nov 5, 2026.
What should Teleport alternatives cover?
Any privileged access management tool for infrastructure has to answer the same questions. We applied these criteria to both columns of the table above, and they work for any Teleport alternatives on your shortlist:
Resources: which protocols it proxies beyond SSH and Kubernetes.
Credentials: whether engineers and agents ever hold the upstream secret.
Controls inside the session: whether it acts on a single query or response.
Audit and just-in-time access: what gets recorded, and how time-bound exceptions are approved.
AI agents: how an agent gets an identity tied to the human who started it.
Give agents scoped access and control the data they receive
An agent can move from a model request to a tool call, a direct API request, and a production query. Formal applies policies to supported traffic across that workflow.
Keep upstream secrets out of the agent
Formal can inject configured database credentials, provider keys, API tokens, and remote MCP OAuth credentials at the proxy.
Limit what the agent can read and change
Policies can block destructive database operations, mask sensitive results, and restrict supported remote tool and API actions, using the agent identity and request context.
Trace actions back to an identity
Formal records supported queries, commands, tool calls, and model activity. Agent identities and human attribution help security teams investigate what happened.
Secure the full agent workflow
Teleport provides infrastructure identities, MCP permissions, and a Beams runtime with an LLM proxy. Formal focuses on governing supported agent traffic across databases, HTTP APIs, model providers, and remote MCP servers.
The comparison covers how an agent receives identity, obtains temporary access, uses credentials, calls models and tools, and reaches production systems. It also distinguishes protocol-level controls from the gateway, network, or runtime integrations Teleport provides.
Limit what the agent can retrieve and send
A coding agent might read a production table, call a remote tool, and send the result to a model provider. A useful policy needs to account for the data returned as well as the action requested.
Formal can allow a supported read operation, mask sensitive fields in the result, block a destructive statement, and apply separate policies to the next supported model or remote tool request. Each connection must pass through the relevant Formal proxy for those policies to apply.
Why teams move from Teleport to Formal
Scope access to the agent task. Give an agent identity-based, time-bound access and apply protocol policies to the requests it makes within that access.
Reduce sensitive data in model context. Mask supported database results and model or tool content before protected information moves to the next service.
Cover direct connections as well as tools. Agents also use database clients, HTTP SDKs, and subprocesses. Endpoint can inspect supported network traffic selected by routing rules outside a single MCP client.
Teleport (goteleport.com) gives engineers certificate-based access to servers, Kubernetes and databases, and sells a privileged access management (PAM) offering on top. Teams look for Teleport alternatives when the question shifts from who can connect to what each query is allowed to return, and when AI agents start opening those connections. We build Formal, so read this as one vendor’s view: every Teleport fact links to Teleport’s own pages.
Where does Teleport fit well?
Both are reasonable choices. They fit different needs.
Teleport positions its PAM offering as “Vault-free Privileged Access” with just-in-time access (Teleport homepage). Here is where each product fits, from each vendor’s own pages.
Choose Formal if
Policy has to act on what a session returns: mask a column in a Postgres response or block a DELETE in production (enforcement docs).
AI agents reach production, and each agent query should be logged with the person who started it (OIDC docs).
Coding agents on laptops and in sandboxes need network rules too: Formal Endpoint applies them to Claude Code, Cursor and Codex traffic (network rules docs).
Choose Teleport if
Your estate includes GitHub: Git repositories are listed as Teleport Protected Resources.[4]
You need moderated sessions, where a moderator can watch a session in real time, take part in it and end it.[9]
See the difference
Same client, same query. The Connector changes what comes back.
Masking query results, not only granting tables
Teleport documents database access by labels, database users and names, plus table-level permissions for PostgreSQL (Database Access Controls); masking result data is not documented. Formal evaluates policies at session, request and response stages, so a policy can mask, filter or block rows as they come back. Our data masking page goes deeper.
Masks the username of every column labeled as an email address. From the Policy enforcement docs.
An agent's DELETE in production
An agent’s query goes through the same request policy as an engineer’s, so a DELETE against production is stopped before it reaches the database. Our Devin guide sets this up end to end: the agent’s OIDC token becomes a Formal machine user, and the log names the human who launched it as the end user (OIDC docs).
Connector log, illustrative
formal connectorproduction
AGENT devin (machine user), end_user: alice@example.com
DELETE FROM orders WHERE created_at < now();
BLOCK DELETE operations are not allowed in production
# logged with both identities
Show the policy
package formal.v2import future.keywords.if# Block all DELETE statements in productionrequest := { "action": "block", "type": "block_with_custom_message", "message": "DELETE operations are not allowed in production", "reason": "Production data protection policy"} if { input.resource.environment == "production" input.sql_query.statement_type == "DELETE"}
Teleport is built around identity and short-lived certificates. Formal is built around a protocol-aware proxy that reads each request and response and enforces policy on it. Both cover SSH, Kubernetes and databases, with the control point in different places.
Architecture
Where the control point sits
Engineers and agents connect to a Connector that runs in your VPC, and it reads each query before the resource does. The Control Plane we host can’t interact with your data.
vs. Teleport:Teleport Agents proxy traffic to your resources and check the roles encoded in each user’s certificate.[2]
connector: postgres-prodillustrative
HUMAN alice@example.com psql
ALLOW SELECT id, email FROM users
MASK email (response stage)
# the Control Plane never sees these rows
How it works
In Teleport, Teleport Agents proxy traffic to your resources and check the roles encoded in each user’s certificate, signed by the Teleport Auth Service (Teleport architecture). In Formal, the Connector is a reverse proxy you run as a container in your VPC. It interprets the protocol, identifies the user, and evaluates policies against requests and responses. The Control Plane is hosted by Formal and cannot interact with your data (architecture docs).
Credentials
Credentials stay on the Connector
Nobody on your team holds the database password, the SSH key or the kubeconfig. The Connector signs in upstream with a Native User, and people keep psql, ssh and kubectl.
vs. Teleport:users sign in with short-lived client certificates signed by the Auth Service.[2]
connector: app-01illustrative
HUMAN alice@example.com ssh app-01
INJECT native user: deploy
ALLOW session opened and recorded
# the Connector holds the key; alice never sees it
How it works
A Native User defines how the Connector authenticates upstream: a password, an SSH key (optionally CA-signed), an AWS, GCP or Azure identity, or a kubeconfig. People and agents sign in as themselves and never handle these secrets (Native Users docs).
AI agents
AI agents and the human behind them
An agent gets its own identity at the proxy, and its queries carry the person who started it. Response policies apply to what the agent gets back, and the MCP gateway logs every tool call.
vs. Teleport:MCP server proxying and per-tool audit in every edition; per-agent certificates and agent-to-owner linkage in Enterprise editions.[3]
connector: agent sessionillustrative
AGENT github-actions (OIDC), short-lived Formal credential
HUMAN end_user: alice@example.com
MASK email column in the Postgres response
ALLOW MCP tool call, logged by the MCP gateway
How it works
Teleport offers MCP server proxying with per-tool audit events in every edition, and per-agent certificates in its Enterprise editions (feature matrix). The same matrix lists per-session MFA for sensitive agent actions and recording of agent sessions. Its Secure MCP page describes short-lived privileges issued per request, with every query attributed to an identity and logged. Those controls decide whether a query runs; none of the Teleport pages we read describe masking or rewriting what it returns. Its Agentic Identity Framework is a set of reference architectures and SDKs; Teleport states it is “not a licensed product or feature”.
Formal treats an agent like any other identity at the proxy. The Endpoint exchanges OIDC tokens from AWS, Azure, GCP, GitHub Actions or SPIFFE for short-lived Formal credentials (OIDC docs). On an OIDC machine session, policies see both the machine user and the human it runs for (end_user). The MCP gateway on the Connector signs MCP clients in, adds the upstream credential and logs every tool call. For Claude Code and Cursor on laptops, see AI coding agent security; the AI agent identity guide covers the concept.
Sessions and policy as code
SSH and Kubernetes sessions
Session recording covers every SSH session and every kubectl exec and attach session, and a Rego policy can require MFA before the shell opens. Policies ship with Terraform, Pulumi or the Kubernetes operator.
vs. Teleport:recording with playback and live view for SSH, Kubernetes, desktops and databases; session moderation in Enterprise editions.[3]
connector: k8s-prodillustrative
HUMAN alice@example.com kubectl exec api-7f9c
VERIFY mfa required (session stage)
ALLOW exec session opened and recorded
# risk score from 1 to 5 in the Sessions app
How it works
The Connector records every SSH session and every kubectl exec and kubectl attach session, logs every other Kubernetes API request, and runs Kubernetes policies on each one. The policy below, from the Kubernetes docs, requires MFA for kubectl exec. Setup details live on our SSH and Kubernetes pages.
Why do teams pick Formal as a Teleport alternative?
Teams pick Formal when access control has to reach inside the session: masking a column in a Postgres response, blocking a DELETE in production, requiring MFA for kubectl exec, or limiting what a cloud agent can query. The same policy engine covers engineers, machine users and AI agents.
At Cursor, Formal Connector took over from access proxies the team had built in-house, and it now governs how engineers and AI agents reach SSH, PostgreSQL, Redis, ClickHouse and Kubernetes (Cursor case study). At Notion, the security team ran a vendor bakeoff before choosing Formal (Notion case study).
What changes for engineers when you switch from Teleport?
Engineers keep their clients. The hostname changes.
Do engineers have to change clients?
No. They point psql at the Connector hostname, ssh to the Connector with a resource name, and kubectl through a kubeconfig the Formal Endpoint updates. VS Code and Cursor Remote SSH, SCP/SFTP and rsync work too.
Through the Connector, which you run in your own VPC. The Control Plane, hosted by Formal, can’t interact with your data; only the Connector and the Endpoint can.
The Connector runs on Linux, AMD64 or ARM64. We recommend AWS ECS Fargate or Kubernetes with Helm (EKS, GKE, AKS or on-premises), and AWS PrivateLink to the Control Plane is available.
Can Formal run next to Teleport during the switch?
Yes. The Connector is a separate proxy with its own hostname, so Teleport keeps serving every resource you haven’t moved yet, and you move one team and one resource at a time.
One resource at a time, with Teleport still running.
Deploy the Connector next to Teleport
Run it with Helm on Kubernetes or on ECS Fargate, in a network that already reaches your resources. Teleport keeps serving everything you haven't moved.
Recreate roles as Rego policies
Map each Teleport role to Formal policies and Native Users, and manage them with Terraform or Pulumi. Decide which access should become a Slack-approved exception instead of a standing role.
Move resources one at a time
Point one team's psql, ssh or kubectl at the Connector, check the logs and session recordings, then move the next resource.
Teleport pricing is usage-based, and Teleport is only partly open source. The Community Edition code is on GitHub under AGPL-3.0, but its binaries are commercial with restrictions: free for companies with fewer than 100 employees and under US$10M in annual revenue. Both Enterprise editions are commercial (feature matrix).
Enterprise is billed on two metrics, per Teleport’s pricing guide (revision June 16, 2026):
Monthly Active Users (MAU): any unique interactive user who shows activity at least once in a monthly period.
Teleport Protected Resources (TPR): any unique resource (Kubernetes cluster, SSH server, database instance, desktop, serverless endpoint, MCP server, Git repository, or machine and workload identity) registered with the cluster in a monthly period, averaged hourly.
The Standard bundle includes Zero Trust Access, Machine & Workload Identity and Identity Governance. Premium adds Identity Security. Teleport’s pricing page shows a “Get Pricing” form rather than list prices.
How is Formal priced?
Formal pricing is scoped to the resources you protect and the people and agents that use them, aligned with how you grow and competitive with the incumbents.
Which other Teleport competitors should you compare?
If StrongDM is also on your list, note that it’s now part of Delinea, which completed the acquisition on March 5, 2026. Our roundup has all 10 PAM tools compared; here is the short version for these three, from each vendor’s own pages:
Teleport
StrongDM
Formal
Core idea
Short-lived certificates and the roles in them
Runtime authorization, sold inside the Delinea Platform
A proxy in your VPC that reads each query and response
Partly: the Community Edition source is AGPL-3.0, while its binaries and both Enterprise editions are commercial, per Teleport’s feature matrix. The pricing section lists the company-size limits.
Is Teleport free to use?
Community Edition is, within its company-size and revenue limits. Teleport lists a free trial for Enterprise (Cloud). What Community Edition leaves out matters for production: Device Trust, session moderation, and access request workflows beyond the CLI, per the feature matrix.
How is Teleport priced?
On two usage metrics, Monthly Active Users and Teleport Protected Resources, in Standard and Premium bundles with no published list prices, per the pricing guide. The pricing section defines both metrics.
How is Formal different from Teleport for AI agents?
Teleport extends certificate-based identity to agents and MCP servers. As a Teleport alternative, Formal applies the same policies to agent traffic as to engineers: masked columns in a Postgres response, MCP tool calls through the MCP gateway, and the human behind an OIDC machine session visible to policies.
Does Formal work with native clients like psql, ssh and kubectl?
Yes. psql, ssh and kubectl point at the Connector, and VS Code or Cursor Remote SSH, SCP/SFTP and rsync work too, per the SSH docs. What changes for engineers covers the rest of a switch.
Who does Teleport compete with?
Teleport competitors for infrastructure access include StrongDM (now part of Delinea), CyberArk (now Idira at Palo Alto Networks), BeyondTrust Privileged Remote Access, HashiCorp Boundary and Formal. Compare Teleport alternatives on where the control point sits, who holds the upstream credential and what each enforces inside a session. Our StrongDM, CyberArk and BeyondTrust pages cover infrastructure access, session controls, and agent identity, and our SSH and Kubernetes pages cover setup.
Is Teleport a proxy?
In part. Teleport’s control plane is the Auth Service plus the Proxy Service, and Teleport Agents in your infrastructure proxy traffic between users and protected resources (Teleport architecture). Access follows the roles in each user’s certificate. Formal’s Connector is a protocol-aware reverse proxy that also reads each query and response.
Teleport or Tailscale: which fits infrastructure access?
They work at different layers. Tailscale describes itself as a secure networking solution that makes encrypted point-to-point connections with the WireGuard protocol, so only devices on your private network can talk to each other (What is Tailscale?). Teleport gives identity-based access to servers, Kubernetes and databases through its Proxy Service and Agents. Formal reads each SQL query, SSH command or Kubernetes request and applies policy to it.
Can Formal mask data in database query results?
Yes. Response-stage Rego policies mask columns as rows return through the Connector: nullify, hash, fake data, constant or partial redaction. The client runs its normal query. See the enforcement docs and our data masking page.
How does Formal compare to Teleport for AI agent security?
Teleport provides infrastructure identities, MCP permissions, and a Beams runtime with an LLM proxy. Formal focuses on governing supported agent traffic across databases, HTTP APIs, model providers, and remote MCP servers. Compare the controls for each stage of your agent workflow in the table above.
Can an agent use production systems without holding their secrets?
Yes. Formal can broker configured upstream credentials at the proxy, so an agent can use a database, API, model provider, or remote MCP service without holding that upstream secret. Pass-through authentication is a separate mode that forwards a client-held credential.
What if an agent calls an API or database directly instead of using MCP?
Formal Endpoint can intercept supported requests selected by network rules, including requests from scripts and native database clients. Formal Connector applies the relevant protocol policies. Coverage depends on the routing, client TLS trust, and endpoint enforcement configuration.
Can we give an agent fewer permissions than its human operator?
Yes. Formal policies can distinguish supported agent identities from their human operators. For example, an agent can be allowed to read approved tables or invoke a read-only remote tool while write operations are blocked.
How does Formal address prompt injection?
Formal Insights analyzes recorded LLM traffic for risks such as prompt injection. Inline protocol policies separately control supported tool, database, and API actions and can mask or block sensitive content, limiting what an agent can retrieve or send.