English | 한국어
edc is everyday carry for systems work: a terminal toolkit for SEs and SREs to inspect hosts, follow resource use, investigate failures, and run remote tasks.
Start with the host: info shows its configuration and current state, and top follows CPU, memory, disk, network, and processes. From there, inspect disk usage, check a connection, trace activity, or record what happened for later comparison. Linux and macOS share the CLI; individual commands have platform and privilege requirements.
The screen uses synthetic demo data with the current UI renderer. Reproduce the captures.
| Task | Start here | Details |
|---|---|---|
| Inspect a host and its resource use | edc info, edc top |
Host information, dashboard |
| Find what is broken now | edc check, edc check --group <group> |
Health check, check routines |
| Find a file by name | edc find, edc find nginxconf /etc |
Find files |
| Find where disk space goes | edc du /var, edc clean |
Directory usage, known caches |
| Investigate a service or container | edc svc, edc ctr |
Services, containers |
| Look into one process | edc proc <pid|name> |
Process view |
| See scheduled jobs and their last result | edc jobs |
Scheduled jobs |
| Check a connection | edc dns lookup example.com, edc tcp check example.com:443 |
Network checks, combined diagnosis |
| Trace application and kernel activity | edc trace |
Tracing and packet capture — support varies by trace |
| Record and compare runs | edc top -w, edc history, edc log -- <command> |
Resource history, command logs |
| Work across hosts | edc remote, edc clean --group <group> |
Remote recipes, remote disk survey |
Inspection commands do not automatically repair the host. Explicit actions have effects: remote runs recipe commands, check <routine> runs the commands of a configured routine, watch fs --exec and --rules run configured actions, and route, disk, and change modify host settings. Recording, log rotation, pruning, setup, and updates also write or delete data. compress and decompress write files, and their --rm deletes the input. See each command's requirements before using it in automation.
Install · Quick start · Configuration · Build from source · Exit codes
Linux and macOS releases include amd64 and arm64 binaries. The install script selects the binary and verifies its SHA-256 checksum.
curl -fsSL https://github.com/x-mesh/edc/releases/latest/download/install.sh | shInstall paths, PATH setup, updates, and source builds →
After installation, check the version and inspect the host without a public IP lookup:
edc version
edc info --public=falseinfo prints system, memory, network, disk, and process information. Open the live resource dashboard next:
edc topPress q to exit. For bounded output or a script, collect five samples as JSON lines:
edc top --count 5 --json ---json - writes JSON to stdout. --json report.json writes it to a file. If --json is the last argument or the next argument is another option, it also writes to stdout. An argument such as -report.json, which starts with - and has a . or a /, is a file name.
Use the task table above to choose the next command. edc help lists commands, and edc help top shows a command's options. Source builds use ./bin/edc in place of edc.
Inspect configuration, memory, disks, and listening ports with info. On Linux and macOS, check looks for failed services, defunct processes, full disks, memory kills, and other current problems in one run. On Linux, inspect systemd services, recent changes, and kernel events. On Linux and macOS, inspect Docker containers through a Unix socket.
edc info --public=false
edc check
edc svc
edc ctr
edc changes --since 2hHost information, health check, service and container views, audit, and events →
By default, edc ctr uses Docker Engine API data only. See the container guide for the optional local Linux collector.
Follow host resources and the busiest processes in top. Switch views to inspect CPU, memory, disk, or network activity; use process filters to focus on a service. ai shows local Claude Code and Codex token usage and account limits. ai sessions shows which sessions need you, and ai agents shows the Claude Code sessions, subagents, and background shells that run now.
edc top --process postgres
edc aiDashboard controls, process filters, and AI usage →
du quickly finds where disk space goes. It scans directories with 4 workers by default and lets you explore the largest entries first.
clean is a read-only disk survey. It does not delete files. It measures known caches and other candidates and shows cleanup commands when requested; you choose whether to run them.
edc du /var
edc cleanThe screen uses synthetic demo data. Both commands support Linux and macOS; some filesystem and Docker details are platform specific.
compress and decompress make an old log smaller on a host without zstd or pigz. They write .gz or .zst on several cores. They keep the input unless you add --rm, and they do not overwrite a file. See Compress and decompress files.
Disk usage and cache inspection details →
Record resource observations, browse earlier runs, and compare metrics over time. Snapshots and saved reports help compare host state; log keeps command output and execution history.
edc top -w
edc history trend
edc log -- ls -lhStop top with q, then open history trend to inspect the recording. The screen illustrates a multi-day history with synthetic data; a new recording starts with the samples you collect.
Recording, trends, growth, and command history details →
Follow application requests and kernel activity to investigate a failure. HTTP trace shows request paths, response status, and measured latency; press i to preview a message below the event list.
Screen rendered with synthetic demo data. HTTP trace requires Linux 5.15 or later with eBPF privileges, or macOS with packet capture privileges. On a host with the required capabilities, trace a local HTTP service on port 8080:
sudo edc trace http --side server --port 8080Press Ctrl-C to stop collection and print the summary. Trace support varies by OS, kernel, and protocol.
HTTP tracing details → · All traces and packet capture →
Use inventory groups and recipes to plan work across hosts, execute steps, and verify results. OpenSSH supplies connection settings. Recipes execute the commands you configure.
edc remote --list
edc remote --dry-run --group <group> --recipe <recipe.yaml>Configure an inventory and recipe before running these examples. The demo shows a plan using example files; it makes no SSH connections. --dry-run displays planned work without executing recipe commands.
Inventory, recipes, execution, and SSH tunnels →
Check a specific layer, or use doctor for a combined diagnosis:
edc dns lookup example.com
edc tcp check example.com:443
edc tcp scan --ports 22,80,443 example.com
edc doctor https://example.comDNS, TCP, TLS, HTTP, doctor, quality, and file watching details →
Watch a directory and run a command when a file changes:
edc watch fs ./src -r --event create,modify --match '**/*.go' --exec 'go test ./...'
edc watch fs --rules watch.yaml --dry-run--rules reads the rules from a YAML file. To keep the watch active after logout, run it as a systemd service.
File watching, rules files, and service setup →
On Linux, explicitly manage routes, grow disks, or apply protected access changes with rollback. Review the privilege, confirmation, and recovery requirements before applying a change.
For a route switch with an active rollback timer:
flowchart LR
A["Save current route<br/>Arm rollback timer"] --> B["Switch exit"]
B --> C{"Verify and confirm"}
C -->|"Confirmed"| D["Disarm timer<br/>Keep new route"]
C -->|"Failure or cancel"| E["Restore saved route"]
A -. "Timer expires without confirmation" .-> E
Route switching, protected changes, and disk growth →
| Guide | Contents |
|---|---|
| Installation and configuration | Installation, updates, build, language, defaults, completion |
| Host inspection | Host information, health check, services, containers, audit, changes, events |
| Live resource monitoring | Top controls, thresholds, processes, AI usage |
| Network checks | DNS, TCP, TLS, HTTP, doctor, location, file watching |
| Disk usage and known caches | Directory browser, caches, Docker, worktrees, remote surveys |
| Recordings, history, and reports | Recordings, snapshots, trends, growth, reports, command logs |
| Remote tasks and SSH tunnels | Inventory, recipes, execution plans, SSH tunnels |
| Protected host changes | Routes, rollback, disk growth |
| Tracing and packet capture | Application protocols, kernel traces, payloads, PCAP |
0: success, or warnings only1: one or more probes fail, orreport difffinds a worse probe2: a run error, such as an argument, a config, a log start, or a report parse error3: not enough privilege for a privileged task4: user cancel, which includes a cancelled selection and Ctrl-C inremoteanddoctor124: the finallogattempt exceeds its timeout
| Capability | Platforms |
|---|---|
| Host resources, directory usage, cache inspection, network probes | Linux and macOS |
| Host audit | Linux and macOS (macOS: some checks) |
| systemd, recent host changes and kernel events | Linux |
| Application and kernel tracing | Varies by trace, kernel, and privileges |
| Remote recipes | Uses local OpenSSH; remote commands depend on each host |
Platform requirements and external lookups · Trace requirements · Configuration and language
MIT. See LICENSE.





