Technology
Claude Code
Monitor Claude Code with cost visibility, adoption tracking, and productivity signals across your engineering teams — without changing how developers work.
Overview
Claude Code
Claude Code is Anthropic's agentic coding tool that runs in your terminal. It can emit OpenTelemetry metrics and events about every session — token usage, cost, tool calls, and the pull requests it helps ship. Sending that telemetry to Dash0 gives engineering leaders and platform teams a single place to understand how Claude Code is adopted, what it costs, and the impact it has — without changing how developers work.
Use Cases
- Cost visibility: Track spend by user, team, and model so you can understand and control what Claude Code costs.
- Adoption tracking: Follow active users, new adopters, and usage distribution to see how Claude Code rolls out across your engineering teams.
- Productivity signals: Correlate sessions, tool calls, and assisted pull requests to understand the impact Claude Code has on your team's output.
Setup
Overview
The Dash0 agent plugin captures Claude Code agent activity as OpenTelemetry traces: tool calls, LLM invocations, token usage, and errors. Send that data to Dash0 and the AI Coding section reports what Claude Code costs, how widely your teams use it, and what it helps ship.
Prerequisites
- Claude Code installed and authenticated on macOS, Linux,
or Windows, on
amd64orarm64. - On Windows, Git for Windows
- A Dash0 auth token with ingest permissions.
- Your name and email set in your git configuration. The plugin uses them to attribute activity to users:
git config --global user.name "{{user_full_name}}"
git config --global user.email "{{user_email}}"Installation
Install the plugin from the official Claude Code marketplace by running this command inside Claude Code:
/plugin install dash0@claude-plugins-officialWhen asked for the installation scope, choose the user scope. It installs the plugin once for every project you work on, and it covers git worktrees and multiple clones of the same repository.
Configure
Configure the plugin through the plugin UI: run /plugin, select dash0 under Installed, choose Configure,
and set the OTLP endpoint, auth token, dataset, and team name. The auth token goes into your OS keychain. Run
/reload-plugins afterwards to apply the changes.
Alternatively, ask the agent directly by entering a prompt like this:
Configure the Dash0 plugin and set the following values:
OTLP-Endpoint: {{endpoint_otlp_http}}
Token: {{token_ingestion}}
Dataset: {{dataset}}
Team: <your team>Verify
- Restart Claude Code. On session start you should see
dash0: connected. - Run any prompt. Within a few seconds, the telemetry appears in the AI Coding section of Dash0.
Organization-wide deployment
Instead of having every developer install and configure the plugin, admins can roll it out to the whole organization using Claude Code managed settings. Managed settings take top precedence, and user or project settings cannot override them. This requires a Team or Enterprise plan and the Owner or Primary Owner role.
1. Install the plugin on members' machines
In the claude.ai Organization plugins admin page, set the plugin to Required (auto-installed, not removable) or Installed by default. This is what downloads the plugin onto each member's machine, and it takes effect on their next session.
Managed settings only enable and configure a plugin. They never fetch it. Skipping this step is the most likely reason
a rollout produces no telemetry: enabledPlugins names a plugin that was never downloaded, so there is nothing to
load.
2. Push the configuration
Add the following payload in claude.ai → Admin Settings → Claude Code → Managed settings. Clients fetch it at startup and re-poll hourly.
{
"$schema": "https://json.schemastore.org/claude-code-settings.json",
"enabledPlugins": { "dash0@claude-plugins-official": true },
"pluginConfigs": {
"dash0@claude-plugins-official": {
"options": {
"OTLP_URL": "{{endpoint_otlp_http}}",
"AUTH_TOKEN": "{{token_ingestion}}",
"DATASET": "{{dataset}}"
}
}
}
}Providing AUTH_TOKEN here makes the rollout fully zero-touch, but the token then sits in plaintext on every machine in
the fleet (in ~/.claude/remote-settings.json) and in the console. Use an ingest-only token scoped to the single dataset
you send to, so a leak can write telemetry there and read nothing. If policy forbids plaintext credentials at rest, omit
AUTH_TOKEN and have each developer add it once via /plugin → Configure, which stores it in the OS keychain. That
costs you the zero-touch rollout.
Do not let developers also self-install the plugin. If a developer already installed it at user scope, both installs are enabled and every span is exported twice. Let managed settings be the single source of truth.
3. Keep team names out of managed settings
Do not add TEAM_NAME to the managed settings payload above. Developers cannot override a managed value, so setting it
centrally tags every developer with the same team and leaves nobody able to correct it. Omitting it is safe:
pluginConfigs merges per key, so the rest of the managed configuration still applies while each developer's own
TEAM_NAME passes through.
Per-developer team names come from team_name: in ~/.claude/dash0-agent-plugin.local.md, or from the DASH0_TEAM_NAME
environment variable for the Cursor, Copilot, and Codex integrations. Distributing those per-user values is an
organizational concern the plugin cannot solve. Whichever system already tracks team membership has to write the file or
export the variable, and update it when someone changes teams. A repository-level .claude/dash0-agent-plugin.local.md
takes precedence over the user-level file, so a checked-in team name file hides a provisioned one for anyone working in
that repo.
To verify the rollout, a developer can run /status → Setting sources (should list Enterprise managed settings) and
claude plugin list (should show exactly one Dash0 identity with Status: ✔ enabled and Scope: managed).
For the full managed-settings reference, including the Dash0 marketplace payload, the MDM on-disk alternative, per-team attribution, and network prerequisites, see the Dash0 Agent Plugin README.