The core difference: reactive vs. preventive
Teleport is excellent at access management: certificate-based authentication, role-based access policies, session recording, and audit logs. If you want to know who connected to prod-server-1 at 2:43 AM and what their session looked like, Teleport gives you that.
What Teleport doesn't do is stop a command from executing. Once a user or agent has access to a server, they can run anything within their OS-level permissions. Teleport will record it; it won't prevent it.
DROP TABLE users or rm -rf /var/data. You'll have a perfect recording of it happening. expacti would have blocked it and asked you first.
These tools are complementary, not competitive. Teleport handles identity and access; expacti handles per-command gates. Teams with strict security requirements often use both.
| Capability | Teleport | expacti |
|---|---|---|
| Access control | ||
| Certificate-based SSH auth | ✓ | via SSH CA |
| RBAC for server access | ✓ | ✓ |
| Per-command approval gateBlock execution until human approves | – | ✓ |
| Whitelist engineExact, glob, regex patterns | – | ✓ |
| Visibility | ||
| Session recording | ✓ | ✓ |
| Audit log | ✓ | ✓ |
| Real-time reviewer viewLive terminal feed to reviewer during session | – | ✓ |
| Risk scoring per command | – | ✓ |
| Anomaly detection | Basic | ✓ |
| AI agent support | ||
| SDK for AI frameworksLangChain, Vercel AI, etc. | – | ✓ |
| Designed for automated agentsNon-human principals, approval loops | – | ✓ |
| Setup & cost | ||
| Self-hosted option | ✓ | ✓ |
| Free tier | OSS core | ✓ |
| Setup complexity | Medium–High | Low–Medium |
| Enterprise pricing | $$$ | Custom |
You need to control what an AI agent or automated script does, not just where it can connect. You want a human-in-the-loop for high-risk commands, with a whitelist for routine work.
Your primary need is identity-based access control and certificate management across a large fleet. Teleport's OSS core is excellent for managing human access to infrastructure.
See expacti in action
Interactive demo — no signup, no install. See a command intercepted, reviewed, and approved in real time.
Different layers of the same security stack
Boundary is positioned as a zero-trust access layer: dynamic host catalogs, short-lived credentials, identity-based targeting. It solves the problem of "who can access this resource" and "how do they connect to it."
Like Teleport, Boundary doesn't control what happens inside a session. A user (or agent) that connects via Boundary can do whatever their target-side permissions allow.
If you're in a HashiCorp-heavy stack (Vault, Terraform, Consul, Boundary), expacti adds the missing layer that none of those tools provide: a real-time, per-command approval gate.
| Capability | Boundary | expacti |
|---|---|---|
| Access control | ||
| Identity-based access to targets | ✓ | Via SSH proxy |
| Dynamic host catalogs | ✓ | – |
| Per-command approval gate | – | ✓ |
| Whitelist + TTL rules | – | ✓ |
| Audit & compliance | ||
| Access event audit log | ✓ | ✓ |
| Session content recording | Enterprise only | ✓ |
| SOC 2 compliance reports | – | ✓ |
| AI & automation | ||
| SDK for AI agent frameworks | – | ✓ |
| Approval queue for automated agents | – | ✓ |
| Ecosystem | ||
| Terraform provider | ✓ | ✓ |
| Kubernetes integration | ✓ | ✓ |
| HCP (managed cloud) option | ✓ | SaaS + self-host |
| Free tier | OSS | ✓ |
Add the missing layer to your Boundary setup
expacti plugs directly into SSH-based workflows. Try it in 10 minutes.
Enterprise power vs. developer-friendly simplicity
CyberArk is the right tool for large enterprises with strict regulatory compliance requirements (HIPAA, PCI-DSS, FedRAMP). It has decades of development, extensive integrations, and proven enterprise deployment patterns. It's also expensive, complex to deploy, and built for human privileged users — not AI agents.
The AI agent gap: CyberArk was designed in an era when privileged access meant a human admin logging into a server. It doesn't have a concept of a non-human agent submitting commands for approval. Its session controls are designed around credential vaulting and session isolation, not real-time per-command gates with an LLM-friendly SDK.
For teams that already have CyberArk for human access, expacti handles the AI agent layer that CyberArk doesn't address. They cover different principals with different tooling.
| Capability | CyberArk | expacti |
|---|---|---|
| Core controls | ||
| Credential vaultingSecure storage of secrets, rotation | ✓ | – |
| Session isolationJump host / proxy layer | ✓ | ✓ |
| Per-command approval gate | Limited | ✓ |
| Whitelist-driven automation | – | ✓ |
| AI & modern automation | ||
| SDK for AI coding agents | – | ✓ |
| GitHub Actions integration | – | ✓ |
| LangChain / Vercel AI support | – | ✓ |
| Approval queue with mobile push | – | ✓ |
| Compliance | ||
| Session recording | ✓ | ✓ |
| SOC 2 / ISO 27001 reports | ✓ | ✓ |
| SAML SSO / SCIM | ✓ | ✓ |
| Practicality | ||
| Self-hosted option | ✓ | ✓ |
| Setup time | Weeks–months | Minutes–hours |
| Starting cost | $50K+/year | Free / $29/mo |
| Free tier | – | ✓ |
Enterprise-grade control, developer-friendly setup
Start free, set up in 10 minutes, protect AI agents that CyberArk wasn't built for.
Passive logging vs. active gating
A bastion host is a network architecture pattern, not a security product. It gives you a single entry point (good for firewall rules), SSH traffic logs (useful for audits), and possibly session recordings if you've added tooling like ttyrec or asciinema.
None of that stops a command from executing. The bastion sees the traffic and passes it through. By the time you check the logs, the data is already deleted.
DELETE FROM events WHERE created_at < NOW() - INTERVAL '90 days'. A bastion logs it. expacti blocks it and sends you a push notification with a risk score of 85/100.
If you already have a bastion, expacti slots in as the next layer. You don't need to remove the bastion; you add expacti behind it to gate commands on critical target servers.
| Capability | Bastion host | expacti |
|---|---|---|
| Single entry point (network control) | ✓ | ✓ |
| SSH session logging | With setup | ✓ |
| Per-command approval gate | – | ✓ |
| Whitelist engine | – | ✓ |
| Risk scoring | – | ✓ |
| Anomaly detection | – | ✓ |
| Reviewer dashboard (real-time) | – | ✓ |
| Slack / Teams approval | – | ✓ |
| Mobile push notifications | – | ✓ |
| Compliance reports (SOC2, ISO27001) | – | ✓ |
| AI whitelist suggestions | – | ✓ |
| SDK for AI agent frameworks | – | ✓ |
| Setup complexity | Low (just SSH) | Low–Medium |
| Cost | Server cost only | Free / $29/mo |
Turn your bastion into an active control point
expacti deploys alongside your existing SSH infrastructure in minutes.
How the tools compare at a glance
All of these tools play in the "privileged access" space. The critical distinction is when they intervene in the access lifecycle.
| Feature | Bastion | Teleport | Boundary | CyberArk | expacti |
|---|---|---|---|---|---|
| Core security model | |||||
| Prevents commands before execution | – | – | – | Limited | ✓ |
| Real-time human approval loop | – | – | – | – | ✓ |
| Whitelist-driven automation | – | – | – | – | ✓ |
| Session recording | With setup | ✓ | Enterprise | ✓ | ✓ |
| Audit log | Basic | ✓ | ✓ | ✓ | ✓ |
| Identity-based access control | – | ✓ | ✓ | ✓ | ✓ |
| AI agent support | |||||
| SDK for AI frameworks | – | – | – | – | ✓ |
| GitHub Actions integration | – | Limited | – | – | ✓ |
| Notifications | |||||
| Slack approval buttons | – | – | – | Enterprise | ✓ |
| Mobile push (PWA) | – | – | – | – | ✓ |
| Practicality | |||||
| Free tier | ✓ | OSS | OSS | – | ✓ |
| Setup time | Hours | Days | Days | Weeks+ | Minutes |
The key insight: these tools address different moments
Before access: Boundary, Teleport, CyberArk credential vaulting — control who can connect.
During access: expacti — control what they can do once connected.
After access: All of them (session recording, audit log) — understand what happened.
expacti is the only tool in this list designed for the during layer: active, real-time command interception with a human in the loop. That's the gap the AI agent era created, and that none of the traditional PAM tools were built to fill.
Try the layer that's been missing
Free tier, 10-minute setup. See what it looks like to have a human in the loop for every command.