DevOpsPilot guide

Everything you need to go from sign-up to asking your AI assistant about your servers.

1. Getting started

  1. Create an account and open the link in the verification email.
  2. Go to Servers → Add server, give it a name and hostname.
  3. Copy the install command and run it on the server (see below). It appears online within a minute.
  4. Connect your AI assistant from Connect AI in the dashboard.
  5. Ask: “Check my production servers” or “Which server has the highest CPU?”.

2. Install the monitoring agent

Supported: Ubuntu, Debian, CentOS / RHEL / Rocky / AlmaLinux and Amazon Linux on x86_64 or ARM64, with systemd. You need root (sudo) once, for the install.

curl -fsSL https://devops.chopanekdajainsamaj.com/agent/v1/install.sh | sudo bash -s -- \
  --url https://devops.chopanekdajainsamaj.com --token <one-time token from the dashboard>
  • Add --docker to also report Docker containers (Pro and Business). The agent user joins the docker group; the agent only makes read-only API calls.
  • The installer checks the download’s SHA-256, creates an unprivileged devopspilot user and a hardened systemd service capped at 64 MB of memory.
  • The agent only connects out to DevOpsPilot over HTTPS (port 443). It opens no ports and runs no commands.
  • The token works once and expires after 24 hours; get a new one from the server’s page (“Reinstall command”).

Check it is running, and uninstall:

systemctl status devopspilot-agent
journalctl -u devopspilot-agent -f

sudo systemctl disable --now devopspilot-agent
sudo rm -f /usr/local/bin/devopspilot-agent /etc/systemd/system/devopspilot-agent.service
sudo rm -rf /etc/devopspilot && sudo userdel devopspilot

3. Connect Claude (claude.ai, Desktop, mobile)

  1. In Claude open Settings → Connectors → Add custom connector.
  2. Name: DevOpsPilot. URL: https://devops.chopanekdajainsamaj.com/mcp
  3. Click Connect. Sign in to DevOpsPilot, choose the organization and click Allow.
  4. In a chat, enable DevOpsPilot from the tools menu and ask your question.

No API key needed: Claude signs in with OAuth. Custom connectors depend on your Claude plan.

4. Connect ChatGPT

  1. In ChatGPT open Settings → Apps & Connectors → Advanced settings and turn on Developer mode.
  2. Click Create. Name: DevOpsPilot. MCP server URL: https://devops.chopanekdajainsamaj.com/mcp. Authentication: OAuth.
  3. Sign in to DevOpsPilot and click Allow.
  4. In a new chat, pick DevOpsPilot from the + / Developer mode menu and ask your question.

Menu names change from time to time; look for “Connectors” or “Apps”. Availability depends on your ChatGPT plan.

5. Claude Code, Cursor, VS Code, Claude Desktop (API key)

Create an API key with the mcp scope under API keys, then:

# Claude Code
claude mcp add --transport http devopspilot https://devops.chopanekdajainsamaj.com/mcp \
  --header "Authorization: Bearer <YOUR_API_KEY>"
// Cursor: ~/.cursor/mcp.json
{ "mcpServers": { "devopspilot": { "url": "https://devops.chopanekdajainsamaj.com/mcp",
  "headers": { "Authorization": "Bearer <YOUR_API_KEY>" } } } }
// VS Code: .vscode/mcp.json
{ "servers": { "devopspilot": { "type": "http", "url": "https://devops.chopanekdajainsamaj.com/mcp",
  "headers": { "Authorization": "Bearer <YOUR_API_KEY>" } } } }

6. MCP tools reference

All tools are read-only. They return JSON with the facts, and a note field when data is missing (never a made-up zero).

ToolInputsReturns
list_serversenvironment?, tag?All servers with status and current CPU, memory and disk.
get_monitoring_summaryenvironment?Online / offline / pending counts and the busiest servers.
get_server_statusserverStatus, last report and current readings for one server.
get_cpu_usageserver, range?Current CPU % and average / min / max over the range.
get_memory_usageserver, range?Memory % with used and total bytes.
get_disk_usageserver, range?Root filesystem % with used and total bytes.
get_network_statsserver, range?Bytes per second in and out, plus totals since boot.
get_load_averageserver, range?Load 1 / 5 / 15 minutes, also per CPU core.
get_server_uptimeserverUptime and boot time.
get_system_informationserverOS, kernel, architecture, cores, agent version, tags.
get_docker_containersserver, state?Container state, restarts, CPU and memory (Pro and Business).
compare_server_performanceservers (2-10), range?Side-by-side statistics for several servers.
generate_health_reportenvironment?, range?Facts per server plus threshold-based observations.
get_server_alertsstatus?, server?Open alert incidents (CPU, memory, disk, offline) and recent ones.

server is the server’s name (as shown in the dashboard) or its ID. range is 1h, 6h, 24h, 7d, 30d or 90d, limited to your plan’s history.

7. REST API

Interactive reference: https://devops.chopanekdajainsamaj.com/docs. Authenticate with an API key that has the read scope:

curl -H "Authorization: Bearer <YOUR_API_KEY>" \
  https://devops.chopanekdajainsamaj.com/api/v1/orgs/<ORG_ID>/servers

8. Authentication

  • Dashboard: email and password, short sessions that renew automatically; sign out other devices from Profile & security.
  • API keys: shown once, read-only, scoped (read, mcp), optional expiry, revoke any time.
  • OAuth (Claude, ChatGPT): you approve one organization; the app gets read-only access. Disconnect it any time under Connect AI → Connected apps.

9. Plans and limits

Limits apply per organization: servers, team members, MCP requests per month, reporting frequency and history. See Pricing; your current usage is on the dashboard’s Usage page.

10. Troubleshooting

  • Server stays “Waiting for agent”: run journalctl -u devopspilot-agent -n 50. The server must reach https://devops.chopanekdajainsamaj.com on port 443.
  • “Enrollment token is invalid or expired”: tokens work once for 24 hours. Click “Reinstall command” on the server page for a new one.
  • Server shows Offline: no report for 3 intervals. Check the service is running and outbound HTTPS is allowed.
  • AI says “Monthly MCP request limit reached”: wait until the 1st of next month (UTC) or upgrade.
  • Docker tool says it is not in your plan: Docker monitoring is on Pro and Business; install the agent with --docker.
  • Claude / ChatGPT cannot connect: use the URL ending in /mcp, sign in with the account that belongs to the organization, and click Allow.

11. FAQ

Can the AI change or restart my servers?
No. Every tool is read-only and the agent cannot run commands.
Do you need SSH access or my passwords?
No. The agent sends data out over HTTPS; we never store SSH keys or server passwords.
Can other customers see my data?
No. Each organization’s data is isolated in the application and by database row-level security.
Which features are coming next?
Alerts (email, Slack, Teams), SSL expiry monitoring, deployment monitoring and AI log analysis. See the changelog.