Local-first SSH and AI infrastructure workspace

Your servers.
Your terminal.
Your control.

Manage SSH connections, investigate server issues with context-aware AI, and explore controlled deployment workflows, all from one local-first desktop workspace.

Download Explore Features

Version 1.1.0 for Linux and Windows is public. macOS is not in this release.

Bastion Terminal
Local-first SSH workspace
Connected
  • Contabo Serverroot@10.2.44.18
  • Harbor Gateops@10.8.30.41
  • Scanner Serverdeploy@10.4.18.22
root@workshop-01 · SSH
Contabo Server
root@workshop-01:~$ systemctl status nginx
● nginx.service
  Active: failed (Result: exit-code)
root@workshop-01:~$
Bastion AI
Plan · waiting for you
Low risk

Nginx failed to bind port 80. Nothing has been run.

ss -lntp | grep ':80'
ApproveEditReject
Conceptual illustration. Not a capture of a released build. Host names and addresses are fictional.

The problem

SSH gets you into a server. The real work starts after that.

Investigating errors, understanding logs, reviewing commands, and deploying changes often means switching between terminals, documentation, AI assistants, and scripts.

Bastion is for backend developers, DevOps engineers, SREs, and teams that look after more than one environment, without taking the decision away from the engineer.

Connect. Understand. Operate.

Three jobs, one desktop workspace. These are the jobs in the public 1.1.0 release.

Connect

In 1.1.0

Your infrastructure, organized.

Keep SSH connections, server groups, terminal sessions, notes, and frequently used commands accessible from one workspace.

  • Saved hosts, groups, tags, and search
  • Tabs, splits, and a local shell
  • SSH keys in the OS credential store
  • Snippets and the command palette

Understand

In 1.1.0

Investigate with context-aware AI.

Understand terminal errors, examine logs, and explore potential fixes with AI assistance informed by the relevant server and session context.

  • Ask, Fix, and Plan modes on the active session
  • Terminal and host context attached to the question
  • Proposed commands shown before anything runs
  • Configurable providers: OpenAI, Anthropic, Google, OpenRouter, and an OpenAI-compatible endpoint

Operate

In 1.1.0

Review actions before they run.

Move from an infrastructure problem to a reviewable action plan, with visibility into proposed commands, affected targets, and execution results.

  • Command review on a plan card
  • Explicit approval before a command runs
  • Risk shown on the proposed step

SSH workspace

Manage multiple servers without losing context.

Version 1.1.0 keeps hosts, sessions, and saved commands in one local workspace. Linux and Windows installs are on the download page.

SSH session details

  • Known-host checks

    Confirm a server fingerprint before trusting a new connection.

  • Groups and search

    Separate production and staging, then find a host without scrolling the whole list.

  • Tabs, splits, and snippets

    Keep more than one session open and drop a saved command into the active one.

  • Keys stay in the OS

    Private keys live in the operating system credential store. The interface does not receive the key material.

AI assistant

AI assistance that understands the task, not just the prompt.

Use relevant terminal and infrastructure context to investigate errors, understand commands, and work through troubleshooting decisions.

AI suggests. Engineers decide.

Version 1.1.0 includes Ask, Fix, and Plan. Fix and Plan show commands on a plan card. Nothing in that flow runs until you approve it.

AI assistant

Illustrative workflow

An Nginx service fails to start.

This sequence shows the intended investigation. It is not evidence that every step is automated.

  1. 1. Inspect. The engineer checks the service status in the session.
  2. 2. Share context. Relevant terminal output is supplied to the assistant.
  3. 3. Explain. The assistant describes plausible causes. It does not claim a certain diagnosis.
  4. 4. Propose. The assistant suggests diagnostic commands.
  5. 5. Review. The engineer reads the commands before anything runs.
  6. 6. Decide. Results are inspected, and the engineer chooses the next action.

Human control

Built around visibility and control.

AI suggests. Engineers decide. A proposed command waits for approval before it runs. Deploy verification is still being finished.

  1. 01

    Review

    Read the proposed command, the target host, and the risk before anything runs.

  2. 02

    Decide

    Approve, edit, or reject the plan. The app waits for that choice.

  3. 03

    Verify

    Inspect the output yourself and choose the next step.

Deploy with a plan. Operate with control.

In 1.1.0

Bastion Deploy is in 1.1.0. A plan is built before a run. Rollback pushes a previous Git revision for git-push profiles. It does not restore a whole server, and other sync methods are refused.

Plan → Approve → Run → Verify → Rollback

  • Project-based deployment configuration
  • Git-aware status and diff
  • SSH-based command execution
  • Deployment logs and exit codes
  • Release history
  • Verification and rollback are not guaranteed
Bastion Deploy

Local-first

Local-first infrastructure management.

Bastion is designed to keep infrastructure management centered on your own desktop, with local configuration storage and direct SSH access.

  • Desktop application on Tauri, React, and Rust.
  • Hosts, snippets, and settings in local SQLite.
  • Private keys in the operating system credential store. The interface does not receive the key material.
  • Direct SSH connections from the machine you are using.
  • AI calls go to the provider you configure. There is no claim of fully offline AI.
AI and privacy

Core infrastructure management is designed around local storage and direct connections. When external AI providers are enabled, relevant context may be transmitted to the selected provider according to its configuration and privacy terms.

Read the security model

Building Bastion in the open.

Linux and Windows 1.1.0 are public. The lists below are what that release includes, what is still being finished, and what is planned.

In 1.1.0

  • SSH hosts, groups, tags, and search
  • Tabs, splits, local shell, snippets, and the command palette
  • AI Ask, Fix, and Plan with approval before commands run

Still being finished

  • Bastion Deploy: project plans, SSH execution, logs, and rollback paths
  • Deployment verification
  • AI troubleshooting refinements

Planned

  • macOS installer
  • Team workflows
  • Broader infrastructure integrations beyond SSH
Full roadmap

Questions

What is Bastion Terminal?

Bastion Terminal is a local-first desktop workspace for SSH management, AI-assisted infrastructure troubleshooting, and controlled operational workflows.

Who is Bastion designed for?

Backend developers, DevOps engineers, SREs, sysadmins, and teams responsible for managing remote servers.

Is Bastion an SSH client?

Yes. SSH connection and session management are the foundation. AI assistance and deployment workflows are being built around that foundation.

Does Bastion automatically execute AI-generated commands?

No. Ask, Fix, and Plan show a plan and wait for explicit approval before a proposed command runs.

Does Bastion require cloud infrastructure?

The desktop app stores hosts, snippets, and settings locally and opens SSH connections directly. Optional AI providers send the relevant context you choose to share to that provider.

Which operating systems are supported?

Bastion Terminal 1.1.0 is public for Linux and Windows. macOS is not in this release.

Is Bastion available now?

Yes. Download Bastion Terminal 1.1.0 for Linux or Windows from the download page. The source repository stays private.

Who develops Bastion?

Bastion Terminal is developed by Hexify Technologies Ltd.

Help shape a better infrastructure workspace.

Questions and feedback are sent to hello@bastionterminal.com. This website does not store the message.

This sends your name, email, and message to hello@bastionterminal.com. This site does not store it. See the privacy notice.