Add opt-in WakaTime integration for AI-assisted coding metrics

Open 💬 0 comments Opened Mar 31, 2026 by yuvrajangadsingh

What variant of Codex are you using?

CLI

What feature would you like to see?

Add an opt-in WakaTime integration so Codex sessions can be measured in the same tooling teams already use to evaluate AI-assisted development.

This is less about generic "time tracking" and more about answering practical questions teams already have:

  • how much activity happened in a Codex session
  • what portion was human-authored vs AI-assisted
  • how Codex compares with other tools in existing engineering analytics workflows

WakaTime already solves the collection and reporting side of this. Codex would only need to emit local heartbeats through the user's existing wakatime-cli.

Why this matters

  • Teams adopting AI coding tools want comparable productivity metrics, not just anecdotes.
  • WakaTime is already part of many developers' existing workflow, so this integrates with dashboards they already use.
  • Claude Code already supports WakaTime, so teams comparing agentic tools notice the gap immediately.
  • This can stay fully local and opt-in: no hosted Codex analytics, no new dashboard, no mandatory tracking.

Proposed scope

Keep the feature narrow and low-maintenance:

  • disabled by default
  • configurable in ~/.codex/config.toml
  • invokes the local wakatime-cli in the background
  • sends heartbeats on file reads/writes, with writes marked appropriately
  • rate-limits read heartbeats to avoid noise

Example config:

[wakatime]
enabled = true
# Optional overrides:
# cli_path = "~/.wakatime/wakatime-cli"
# project = "my-project"

This keeps the request concrete:

  • no built-in reporting UI
  • no server-side telemetry requirement
  • no vendor lock-in beyond the user's chosen local CLI
  • just a bridge from Codex file activity to an existing developer analytics tool

Why WakaTime specifically?

The broader need is "make AI-assisted coding measurable in existing workflows." WakaTime is the fastest path to that because it already has:

  • local CLI support
  • project/session attribution
  • existing dashboards and team workflows
  • AI-related metrics support, including --ai-line-changes

If maintainers prefer a more generic activity hook in the long term, that would also solve the same problem. WakaTime is just the most immediately useful starting point.

Are you interested in implementing this feature?

Yes. I already have a working implementation on a fork with opt-in config, a dedicated tracker module, non-blocking background CLI invocation, heartbeat rate limiting, and tests. Happy to open a PR if this aligns with the roadmap.

Additional information

This is a reframed version of #9158 (closed for lack of upvotes). Since then, @alexbadoi raised the same need from a team perspective: measuring human vs machine output. @etraut-openai suggested resubmitting with better framing, so here it is.

cc @alexbadoi @etraut-openai

View original on GitHub ↗