Dash0 CLI

dash0-cli

dash0 is a command-line interface for the Dash0 observability platform. It manages Dash0 assets (dashboards, views, check rules, synthetic checks, recording rules, notification channels, spam filters), queries telemetry (logs, spans, traces, metrics, failed checks), sends telemetry via OTLP, and manages organization entities (teams, members, profiles).

Prefer dash0 --agent-mode <command> --help over guessing flags. Every command's exact, always-current flag list, aliases, and examples are available as structured JSON via --agent-mode <command> --help (e.g. dash0 --agent-mode dashboards list --help). This bundle deliberately does not duplicate flag tables — they'd go stale the moment a flag is added or renamed. Use the topics below for concepts, YAML formats, and workflows that --help output can't express; use --agent-mode <command> --help for the exact flags to pass.

Command taxonomy

Prerequisites

Every command that talks to the Dash0 API or OTLP endpoint needs credentials, resolved in this order (first match wins): environment variables (DASH0_API_URL, DASH0_OTLP_URL, DASH0_AUTH_TOKEN, DASH0_DATASET), CLI flags (--api-url, --otlp-url, --auth-token, --dataset), then the selected profile (--profile flag → DASH0_PROFILE env var → the active profile on disk). See the config and login topics for profile management and OAuth authentication.

Global flags

--api-url, --otlp-url, --auth-token, --dataset, --profile, --agent-mode (env: DASH0_AGENT_MODE), --color (semantic or none), --experimental/-X, --max-retries, --no-skill-hint (env: DASH0_NO_SKILL_HINT — suppress the "start with dash0 skill show … or run dash0 skill install" hint stitched onto errors; dash0 skill show (no arg) prints the entry point with a full topic index, dash0 skill show <topic> drills into one topic without writing files, and dash0 skill install adds a persistent bundle to the project). Run dash0 --agent-mode --help for the full, current list.

Agent mode

Agent mode optimizes the CLI for AI agents: JSON output by default, structured --help, JSON errors, no confirmation prompts, no color. It auto-activates when a known AI agent environment variable is detected (CLAUDE_CODE, CURSOR_AGENT, CODEX, GITHUB_COPILOT, and others), or can be forced with --agent-mode / DASH0_AGENT_MODE=1.

How asset commands work

All seven asset types (dashboards, check-rules, synthetic-checks, views, recording-rules, notification-channels, spam-filters) share the same five subcommands: list, get, create (alias add), update, delete (alias remove). Output formats are table, wide, json, yaml, csv (query commands use table/json/csv only). create/update accept -f <file> (or -f - for stdin) and --dry-run.

dash0 apply -f <file|directory> provides create-or-update semantics across all asset types in one command, and (experimentally, via --since <ref>) delete semantics based on git history — see the apply topic.

Asset identifiers and idempotent upsert

Every asset type accepts a user-defined identifier in its YAML/JSON document. When present, both create and apply perform an upsert (PUT) against that identifier — the asset is created on the first call and replaced on every subsequent call, which is what makes GitOps/CI workflows idempotent. When absent, both perform a plain create (POST) and the server assigns a fresh ID on each invocation, so repeated runs produce duplicates. The identifier field location varies by kind:

Origin vs ID — do not conflate them. Origin (dash0.com/origin label) identifies which system is the authoritative source of truth for an asset (dash0-cli, terraform, ui) — it's provenance metadata, not a lookup key, and the CLI strips it before sending most asset types to the API so it doesn't claim ownership of assets managed elsewhere. ID is the user-defined external identifier used for upsert, described above. Notification channels and spam filters are the two exceptions where origin (not ID) is the upsert key.

When list -o yaml or get -o yaml exports an existing asset, the server-assigned ID is rendered into the correct field, so an export → edit → apply (or update) round-trips through the identifier automatically.

Filter syntax

Query commands (logs query, spans query, metrics instant --filter, failed-checks query) accept --filter expressions in the form key [operator] value (operator defaults to is when omitted), or JSON filter criteria copied from the Dash0 UI. Multiple --filter flags combine with AND logic.

Common attribute keys: service.name, otel.log.severity.number, otel.log.severity.range, otel.log.body, otel.span.status.code, otel.trace.id. Valid otel.log.severity.range values: TRACE, DEBUG, INFO, WARN, ERROR, FATAL, UNKNOWN.

Custom columns

--column (repeatable) selects which columns appear in table/csv output, replacing the default set entirely. Each command has short aliases (e.g. time, severity, body for logs query; duration, status, service for spans query) — see each topic for its alias table — and any OTLP attribute key can also be used directly as a column. Not supported with -o json.

Precision mode (adaptive sampling)

By default the API applies adaptive sampling to log and span queries for speed on large datasets. Pass --precision disabled to logs query or spans query for deterministic, complete results (higher latency) on narrow lookups like a specific trace or request ID; --precision adaptive is the explicit default. traces get always disables sampling and doesn't accept the flag; metrics instant and failed-checks query don't honor it.

Common workflows for AI agents

Set up credentials from environment variables

export DASH0_API_URL=https://api.us-west-2.aws.dash0.com DASH0_OTLP_URL=https://ingress.us-west-2.aws.dash0.com DASH0_AUTH_TOKEN=auth_xxx DASH0_DATASET=default

Then any command picks them up without flags, e.g. dash0 dashboards list or dash0 logs query --from now-1h.

Export an asset, modify it, and re-apply

dash0 dashboards get <id> -o yaml > dashboard.yaml

Edit the file, then either dash0 dashboards update -f dashboard.yaml (ID read from the file) or dash0 apply -f dashboard.yaml (auto-detects create vs. update).

Bulk export and re-apply all assets of one type

dash0 dashboards list -o yaml > all-dashboards.yaml
dash0 apply -f all-dashboards.yaml

The YAML output is a multi-document stream, ready to replay across environments.

Send a deployment event

dash0 logs send "Deployment v2.3.0 completed" \
    --event-name dash0.deployment \
    --severity-number 9 \
    --resource-attribute service.name=my-service \
    --resource-attribute deployment.environment.name=production \
    --log-attribute deployment.status=succeeded

Query errors in the last hour

dash0 logs query --from now-1h --filter "otel.log.severity.range is_one_of ERROR WARN" --limit 200

Non-interactive deletion (for automation)

Always pass --force to skip the confirmation prompt: dash0 dashboards delete <id> --force. With --force, an already-deleted asset is treated as idempotent success — the command exits 0 and prints a short "already deleted" line to stderr. Without --force, a missing asset still surfaces as a non-zero exit with a "not found" error. This is uniform across every delete subcommand and every remove-shaped variant (members remove, teams remove-members); in the multi-target variants the check runs per member so one concurrently-removed member does not fail the whole call.

Validate assets before applying

dash0 apply -f assets/ --dry-run

Sync a directory to match its state as of a git ref (experimental)

apply --since <ref> deletes assets removed from -f's contents since <ref>, in addition to the usual create/update behavior. Requires --experimental/-X and -f's target to be inside a git repository:

dash0 --experimental apply -f assets/ --since HEAD~1 --force

Preview the plan first with --dry-run — see the apply topic for the full reference, including the GitHub Actions invocation pattern and ref-resolution edge cases.

Topics

Run dash0 skill show <topic> for the reference content below, or read references/<topic>.md directly if the skill is installed on disk.

Keeping this skill up to date

If the dash0-cli version this skill shipped with feels stale (missing a command you can see in dash0 --help), run dash0 skill install again to refresh it from the currently installed binary.