alogin — Agent Skill (MCP-first)

Use this skill when the task involves remote server access, health inspection, persistent shell sessions, cluster execution, file transfer, or tunnel management.

Common triggers: SSH, remote shell, server login, bastion, health check, inspect node, run command, persistent session, cluster, SFTP, background job, server memory, alogin.

MCP vs CLI priority

If MCP tools are available in your environment, always prefer them over the
alogin CLI — they carry richer audit metadata, handle reconnection
automatically, and keep the session state server-side.

| Task | MCP (preferred) | CLI (fallback) |
|------|-----------------|----------------|
| Persistent shell | remote_shell | alogin ssh session exec |
| One-shot command | exec_command | alogin ssh connect --cmd |
| Health snapshot | inspect_node | alogin ssh connect --cmd "..." |
| Upload + run script | run_script | alogin ssh run-script |
| Edit remote file | remote_replace | alogin ssh session exec sed |
| File transfer | pushfile / pullfile | alogin scp push/pull |
| Cluster exec | execoncluster | alogin ssh cluster <name> --cmd |
| List servers | list_servers | alogin server list --format json |
| Tunnel manage | starttunnel / stoptunnel | alogin net tunnel start/stop |
| Server memory | getmemory / savememory | alogin agent server-memory list/add |

Standard Workflow

1. DISCOVER  — list_servers / alogin server list --format json
2. RECALL    — get_memory(server_id) before any work on a known server
3. SHELL     — remote_shell(target) → session_id, then reuse for every command
4. INSPECT   — inspect_node before making changes
5. ACT       — remote_shell for sequences; exec_command only for one-shot
6. VERIFY    — read-only check after change
7. SAVE      — save_memory() with findings before closing session
8. CLEANUP   — remote_shell(target, session_id, command="exit")

Always supply an intent parameter on execution tools — it is written to the audit log and is the primary signal for HITL reviewers.

MCP Tool Reference

Query tools (read-only)

list_servers

List servers in the registry.

Returns: [{id, host, user, protocol, devicetype, note, gatewayid}]

get_server

Full detail for one server.

listclusters / getcluster

List cluster groups or get one with full member detail.

inspect_node

Structured health snapshot: CPU load, memory, disk, top processes.

listtunnels / gettunnel

List saved tunnels with live running status, or get one by ID.

get_memory

Retrieve operational notes written by previous agents for this server. Call immediately after identifying the target server.

Returns: [{id, serverid, content, createdat}]

save_memory

Record lasting operational knowledge (PATH quirks, version constraints, service layout). Write entries that will be useful three sessions from now.

Execution tools (write)

remote_shell ⭐ Primary tool

Persistent SSH session identified by session_id. Shell state (cwd, env vars) carries across calls. Reconnects automatically on TCP drop.

// Open session
{"tool": "remote_shell", "arguments": {"target": "3", "intent": "deploy app"}}
// → {"session_id": "abc-123", "status": "created"}

// Reuse session
{"tool": "remote_shell", "arguments": {"target": "3", "session_id": "abc-123", "command": "df -h"}}

// Close session
{"tool": "remote_shell", "arguments": {"target": "3", "session_id": "abc-123", "command": "exit"}}

exec_command

One-shot SSH command. Use only when no follow-up commands are expected.

execoncluster

Run commands on all cluster members in parallel.

run_script

Upload a script string to the remote host, execute it, then delete the temp file. Preferred over pushfile + remoteshell for multi-line scripts.

remote_replace

Safe in-place string substitution in a remote file. Prefer over cat + run_script when editing config files.

Returns: {"replaced": N} — 0 means not found, file unchanged.

Tunnel lifecycle tools

starttunnel / stoptunnel

Start or stop a saved tunnel (tmux-backed).

Always call list_tunnels first to avoid starting duplicates.

CLI Fallback Reference

Use the following when MCP tools are not available.

Server registry

alogin server list --format json
alogin server show <host> --format json
alogin server add --host 10.0.0.10 --user admin
alogin server add --host bastion --user ops --auth-method key --identity-file ~/.ssh/id_ed25519
alogin server alias add prod admin@prod-db
alogin server alias list --format json

SSH sessions (stateful — preferred over one-shot)

# Start session (prints session ID)
id=$(alogin ssh session start web-01)

# Commands share state — cd, exports persist
alogin ssh session exec "$id" "cd /var/log && ls"
alogin ssh session exec --timeout 300 "$id" "make build"
alogin ssh session exec --force-utf8 "$id" "cat /etc/os-release"

# List and reuse
alogin ssh session list
alogin ssh session exec "$existing_id" "uptime"

# Stop when done
alogin ssh session stop "$id"

--timeout N = wall-clock limit (default 30 s). --idle-timeout N = cut if silent for N seconds. --force-utf8 = fix garbled multi-byte output (Korean, Japanese, etc.).

One-shot command (stateless)

alogin ssh connect web-01 --cmd "df -h"
alogin ssh connect user@host --cmd "hostname" --force-utf8

Use only for truly independent, stateless single commands.

Background jobs

id=$(alogin ssh session start web-01)
job=$(alogin ssh session bg-exec "$id" "apt-get upgrade -y")
alogin ssh session job status "$job"
alogin ssh session job logs --follow "$job"
alogin ssh session job list --session "$id" --json
alogin ssh session job cancel "$job"
alogin ssh session job purge

Cluster execution

alogin ssh cluster add web-cluster 10.0.1.1 10.0.1.2
alogin ssh cluster list --format json
alogin ssh cluster web-cluster --cmd "uptime"
alogin ssh cluster web-cluster --cmd "df -h" --format json

File transfer (SCP)

alogin scp push ./deploy.tar.gz web-01:/opt/releases/
alogin scp pull web-01:/var/log/app.log ./
alogin scp push --recursive ./dist/ web-01:/var/www/app/
alogin scp pull -r web-01:/var/log/nginx/ ./logs/

Run local script / inline script

# Local file
alogin ssh run-local ./deploy.sh --remote web-01
alogin ssh run-local ./check.py --remote web-01 --interpreter python3
alogin ssh run-local ./setup.sh --remote web-01 -- --env prod

# Inline / piped content
echo "df -h && uptime" | alogin ssh run-script --remote web-01
alogin ssh run-script --remote web-01 --content "#!/bin/bash\ndf -h\nuptime"

Gateways, profiles, tunnels

alogin net gateway add secure-zone bastion.ext.com internal-jump
alogin net gateway list --format json
alogin server add --host prod-sql --gateway secure-zone
alogin net profile add office --gateway secure-zone
alogin net profile use office
alogin net profile use none

alogin net tunnel add db-proxy --server prod-db --local-port 5432 --remote-port 5432
alogin net tunnel start db-proxy
alogin net tunnel status db-proxy
alogin net tunnel stop db-proxy

Server memory (CLI)

alogin agent server-memory list <server-id>
alogin agent server-memory add  <server-id> --text "nginx config: /etc/nginx/nginx.conf"
alogin agent server-memory del  <server-id> <note-id>

Read server prompt before acting

alogin server show <id> --format json   # includes system_prompt field

The system_prompt field contains server-specific restrictions. Read it before connecting and adhere to it throughout the session.

Safety Rules

• Always provide intent on execution tool calls.
• Read-only inspection before any write operation.
• Do not run destructive commands (rm -rf, shutdown, DROP TABLE) without

explicit user confirmation — HITL policy will block or queue them.

• If a server device_type is router, switch, or firewall, do not assume

standard Linux commands work.

• Check list_tunnels before starting a tunnel to avoid duplicates.
• For remote file edits, prefer remote_replace over full-file rewrites.

Agent Memory Patterns

Three patterns to apply every session:

1. Dynamic profiling — After exploring a new server, call save_memory with

log paths, installed tools, service layout. Next session starts with full context, zero exploration turns.

1. State drift detection — Save baselines (disk %, process counts, config

checksums). On subsequent sessions, compare and flag anomalies proactively.

1. Multi-agent handoff — Notes written by one agent are immediately available

to the next. No duplicate work; no knowledge lost between agents.

JSON Output (CLI)

alogin server list --format json | jq '.[].host'
alogin ssh cluster prod --cmd "uptime" --format json | jq '.[] | {host, output}'
alogin net tunnel list --format json
alogin ssh session job list --json