Documentation
Tink is an AI-powered server mechanic. Install it on any Linux server and it continuously monitors, diagnoses, and guides you through fixes from the CLI.
Overview
Tink sits on your server as a lightweight agent. It watches system health (CPU, memory, disk, network, services), detects issues before they become outages, and explains problems in plain English. When something needs fixing, Tink guides you through the repair step by step from your terminal.
Your dashboard at tink.bot/account shows all your machines, their current health, and a history of every scan and fix. Alerts go to Telegram, email, or any channel you configure.
Installation
One command. Works on any Linux server (Ubuntu, Debian, CentOS, RHEL, Amazon Linux, Alpine).
curl -fsSL https://tink.bot/install | shThe installer downloads the Tink agent binary, creates a systemd service, and prompts you for your account token. The agent starts scanning immediately after install.
Tip: Get your account token from tink.bot/account after signing up. The token links your server to your dashboard.
How It Works
Tink monitors CPU, memory, disk, network, and running services. Scans run every few minutes and adapt frequency based on system load.
When something looks wrong, Tink analyzes the symptoms and explains the issue in plain English. No need to parse log files or decode cryptic error codes.
Tink suggests specific commands to fix the issue. You review each command before it runs. Nothing executes without your approval.
Every scan, diagnosis, and fix is logged. Your dashboard shows a complete timeline of what happened and what was done about it.
Dashboard
Your dashboard at tink.bot/account shows:
- All registered machines with real-time health status
- CPU, memory, disk, and uptime for each server
- Recent scans with issues found and fixes applied
- Alert channel configuration (Telegram, email)
- Account settings, billing, and API tokens
CLI Commands
The Tink agent runs as a daemon, but you can interact with it from the command line:
tink onboardRegister this machine with tink.bottink statusShow current health summarytink reportUpload a scan to the control planetink issuesList current issuestink fixReview and run a supervised fixtink uninstallRemove the agent and systemd serviceTelegram Commands
After you link Telegram from the dashboard, talk to Tink in chat or use these commands:
/statusFleet health snapshot/machinesList machines with health scores/scoreHealth score for every machine, worst first/fleetNumeric fleet snapshot (machines, issues, scans, fixes)/statuspagePublic status page URL to share with customers/whoamiLinked tink.bot account, plan, and machine count/checkShow how to diagnose a machine/check <machine>Immediate diagnosis of one machine/restartShow how to restart a service/restart <service>Restart a service (asks for confirmation first)/hotMachines with the most open issues/trustExecution trust level for every machine/degradedMachines with a health score below 50/alertsLast 10 issues across the fleet/briefOpen critical and warning issues right now/uptime90-day uptime % for every machine/scansScan count in the last 24 hours/fixesFix runs in the last 7 days/failedMachines whose last fix run failed/fixedMachines whose last fix run succeeded/skippedMachines whose last fix run was skipped/nofixMachines that have never had a fix run/summaryLast scan headline for every machine/offlineMachines that stopped reporting/onlineMachines that are currently reporting/seenLast scan time for every machine/digest7-day summary of issues and trends/openCurrently open issues across the fleet/nextHighest-priority open issue/criticalOpen critical issues only/warningsOpen warning issues only/oldestOldest open issues first, with age/staleOpen issues older than 24 hours/infoOpen info issues only/todayIssues opened in the last 24 hours/weekIssues opened in the last 7 days/monthIssues opened in the last 30 days/yearIssues opened in the last 365 days/resolvedLast 10 issues Tink cleared/certsTLS certificates close to expiry/usageCurrent CPU, memory, and disk from the last scan/spareMachines with CPU, memory, and disk headroom/pressureMachines under CPU, memory, or disk pressure/capacityUsed and total memory and disk from the last scan/freeRemaining memory and disk from the last scan/loadCPU load percent from the last scan/idleMachines with idle CPU from the last scan/busyMachines with high CPU from the last scan/diskMachines running out of disk/memoryMachines running out of memory/cpuMachines with high or sustained CPU load/servicesMachines with critical services that are down/portsExposed database and cache ports/sshMachines under SSH brute-force/firewallMachines with the host firewall down/oomMachines with OOM kills or kernel log faults/baselineMachines with unusual spikes vs their own history/forecastResources trending toward failure/streakDays each machine has stayed critical-free/correlateShared issue patterns across two or more machines/channelsConnected notification channels/snoozeShow current mute state/snooze 4hSilence alerts for 4 hours/snooze <machine> 4hSilence one machine only/unsnoozeRe-enable alerts on every channel/quietShow current quiet hours/quiet 23-7Set recurring quiet hours (UTC); warnings stay silent/quiet offDisable quiet hours/oncallShow who's on-call this week/oncall @namePin the current on-call person/oncall rotate a bSet a weekly on-call rotation/ackShow which machines are acknowledged/ack <machine>Mark that you are handling a machine/unackedMachines with open issues that nobody has acknowledged/helpFull command listMCP Server
Connect Claude Desktop, Cursor, or any MCP client to Tink. The server is at https://tink.bot/api/mcp. Send a Bearer token from tink.bot/account in the Authorization header.
get_fleet_statusHealth scores and issue counts for every machinelist_issuesOpen issues, optional machine or severity filterget_machine_detailsMetrics and issues for one machine by nameask_tinkAsk a natural-language question about your fleetAlerts & Channels
Tink sends alerts when it detects issues on your servers. Configure alert channels from your dashboard:
- Telegram - instant messages to your phone
- Email - digest or per-incident alerts
- Webhook - POST to any URL for custom integrations
- Discord - rich embeds with color-coded severity
- Slack - Block Kit messages for your team channel
- ntfy - free push notifications on iOS, Android, and web
- WhatsApp - text alerts via Twilio
- PagerDuty - incidents with auto-resolve on all-clear
Each alert includes the machine name, issue summary, severity, and a suggested fix. Click through to your dashboard for full details.
Rescue Mode
If a server is in bad shape and the Tink agent can't start normally, use rescue mode:
curl -fsSL https://tink.bot/rescue | shRescue mode runs a one-shot diagnostic without installing anything permanently. It collects system info, identifies critical issues, and prints a prioritized fix plan.
Security
- The agent runs as a non-root user with limited permissions
- All communication is encrypted over HTTPS
- Account tokens are scoped per-machine
- Fix commands require explicit user approval before execution
- No shell access is granted to the Tink service
- System data stays in your account; we don't share or sell it
Plans
- Scout (Free) - 1 machine, hourly scans, basic alerts, 50 AI messages/month
- Mechanic ($9/mo per machine) - Unlimited machines, 5-minute scans, supervised fixes, all alert channels, 500 AI messages/month
- Crew Chief ($29/mo per machine) - Unlimited machines, real-time scans, autonomous remediation, unlimited AI messages, priority support
- Rescue ($29 one-time) - Emergency diagnostic session with full autonomous remediation
FAQ
Any Linux distribution with systemd. Ubuntu, Debian, CentOS, RHEL, Amazon Linux, Fedora, and Alpine are all tested.
No. Tink suggests fixes and waits for your explicit approval. Nothing runs without you confirming it first.
Minimal. The agent idles at under 20MB RAM and near-zero CPU between scans. Scans themselves take a few seconds.
Tink is designed for long-running servers, not ephemeral containers. For Kubernetes, point it at your nodes instead.
System metrics (CPU, memory, disk, network), running services, and relevant log snippets when diagnosing issues. No application data, credentials, or file contents.