Dash0 Raises $110M Series B at $1B Valuation
Monitor OpenAI Codex with cost visibility, adoption tracking, and productivity signals across your engineering teams — without changing how developers work.
OpenAI Codex is OpenAI's agentic coding tool that runs in your terminal. With the Dash0 agent plugin it emits OpenTelemetry traces about every session — prompts and responses, token usage, cost, the model behind each turn, tool and MCP calls, sub-agent activity, and the pull requests it helps ship. Sending that telemetry to Dash0 gives engineering leaders and platform teams a single place to understand how Codex is adopted, what it costs, and the impact it has — without changing how developers work.
The Dash0 agent plugin captures OpenAI Codex agent activity as OpenTelemetry traces — prompts and responses, token usage, cost, tool calls, MCP calls, and sub-agent activity, with shared trace context across each turn. Sending that data to Dash0 powers the AI Coding section, giving you cost visibility, adoption tracking, and productivity signals across your teams.
Run the installer:
You are prompted for your Dash0 connection details:
{{endpoint_otlp_http}}{{token_ingestion}}{{dataset}}The installer registers the plugin's hooks in ~/.codex/config.toml as a managed block — preserving any hooks and
config you already have — fetches the codex-on-event binary from
GitHub Releases with checksum verification, and writes your
credentials to ~/.codex/dash0-agent-plugin.local.md (chmod 600).
For more information on choosing a dataset, see the dataset documentation.
Pass the credentials up front to skip every prompt — useful for CI, containers, and fleet rollouts:
The same values can be supplied as the DASH0_OTLP_URL, DASH0_AUTH_TOKEN, and DASH0_DATASET environment variables.
DASH0_TEAM_NAME sets the team name, and DASH0_VERSION pins a specific plugin release instead of the latest one.
If you prefer Codex's native plugin flow:
This path needs two steps the installer would otherwise handle for you: create ~/.codex/dash0-agent-plugin.local.md
with your endpoint and token as shown below, then run /hooks inside Codex and trust the Dash0 hooks by pressing t —
Codex does not auto-trust a plugin's hooks on install.
The configuration lives in the YAML frontmatter of ~/.codex/dash0-agent-plugin.local.md, which the installer writes
for you. Edit it to reconfigure later — changes take effect on the next hook fire, no restart needed:
A per-project .codex/dash0-agent-plugin.local.md inside a repository takes precedence over the global file. The two
files do not merge: if a project-level file exists, the global one is ignored entirely. Set enabled: false to switch
the plugin off for a scope without uninstalling it.
By default the plugin omits prompt content and tool inputs and outputs from spans (omit_io: true) and sends the real
user name and email (omit_user_info: false). Set omit_user_info: true to hash the user name, drop the email, and
redact the working directory.
Codex activity breaks down by user and model out of the box. To also break down by team or department, set team_name
in the configuration file — it is attached to every span as dash0.team.name.
chat <model> span carrying token usage and an execute_tool <Name> span per tool call.If no traces arrive, run /hooks in Codex and check that the Dash0 hooks are listed and Active. Adding
debug: true and debug_file: /tmp/dash0-codex-debug.log to the configuration file logs every emitted span as a
[dash0:trace] line, which tells you whether the problem is span emission or the endpoint and token.
Re-run the installer to upgrade — it fetches the latest release and leaves your credentials untouched. To remove the
plugin, run the uninstaller, which strips Dash0's entries from ~/.codex/config.toml while preserving your own hooks
and config:
Start a new Codex session after either operation.