[Introduction](https://verity.md/docs) [Quickstart](https://verity.md/docs/quickstart) [Authentication](https://verity.md/docs/authentication) [How it works](https://verity.md/docs/how-it-works) [Moments](https://verity.md/docs/moments) [Formal verification](https://verity.md/docs/formal) [Findings & verdicts](https://verity.md/docs/findings) [The Standard](https://verity.md/docs/standard) [Org Standards](https://verity.md/docs/org-standards) [Standards on the web](https://verity.md/docs/web-standards) [The terminal editor](https://verity.md/docs/terminal-editor) [Memory](https://verity.md/docs/memory) [Org knowledge](https://verity.md/docs/org-knowledge) [Token telemetry & cost](https://verity.md/docs/cost) [The dashboard](https://verity.md/docs/dashboard) [CLI & commands](https://verity.md/docs/cli) [Files & configuration](https://verity.md/docs/files) [Security & privacy](https://verity.md/docs/security) [Troubleshooting](https://verity.md/docs/troubleshooting)

Concepts

# Token telemetry & cost

See what your agents spend — per model, per session, per run, per repository, and across everything you can read — so scaling agents never means losing the bill.

Verity meters the token usage of every coding agent working in a repository, from the agent's own telemetry, and shows it in the [dashboard](https://verity.md/docs/dashboard): spend over time, a per-model breakdown, a fleet cost tree, a per-session table, and the cost of any single run. Across repositories, **Token telemetry by org** groups the same figures by the organization that owns each remote. It is the third pillar alongside the [gate](https://verity.md/docs/how-it-works) and [memory](https://verity.md/docs/memory).

Verity itself has a **Free plan** with limits, and paid Pro and Fleet plans. See [pricing](https://verity.md/pricing). The figures on this page are what *your agents* spend with their model providers.

## Enable telemetry

Cost tracking is opt-in, and it is one choice for the whole machine. `verity init` asks once, after you sign in. It does not ask again when telemetry is already on, or when you said no before. You can also turn it on or off at any time from the CLI:

```
verity telemetry install     # turn on
verity telemetry check       # see status
verity telemetry uninstall   # turn off
```

Restart Claude Code after enabling

Telemetry is delivered through Claude Code's OpenTelemetry export, which is configured at startup. After enabling it, restart Claude Code — metrics then appear within about a minute of activity.

## How it flows

`verity telemetry install` writes an OpenTelemetry block into your Claude Code user settings, `~/.claude/settings.json` (`$CLAUDE_CONFIG_DIR` moves it). The block points Claude Code's metrics and traces exporters at the Verity service. Claude Code 2.1.282 and later ignore these settings in a project's `.claude/settings.local.json`, where Verity wrote them before 0.35.0. So the install removes those old keys from the project you run it in.

The block contains a switch, not a secret: the authorization header is supplied at Claude Code startup by `verity telemetry headers`, which resolves your login from `~/.verity/credentials` and is refreshed automatically. Because one login covers every repository you work in, usage is attributed to **the repository the activity came from**, so a session that spans two projects bills each of them correctly.

The block is in your user settings, so it applies to every Claude Code session on this machine. Only Verity projects send data. In any other directory, `verity telemetry headers` exits 1, and Claude Code then exports nothing from that session.

A startup warning outside Verity projects is expected

In a repository that is not a Verity project, Claude Code shows this warning when a session starts:

```
otelHeadersHelper failed (OpenTelemetry export headers unavailable): exited 1: Verity telemetry is off here: this directory is not in a Verity project. Nothing is exported from this session.
```

It means the privacy control worked: nothing left the machine. To stop the warning everywhere, turn telemetry off with `verity telemetry uninstall`.

## What is collected

Telemetry is **metrics and traces only**:

- Token counts and estimated USD cost, by model, session and agent.
- Request timing for each review.
- The names of the tools the agent called, and the repository name.
- The session id, so a run and its cost can be joined to the session it belongs to.

It does **not** include your prompts, your code, or tool input and output — those exporter flags are written *off* explicitly rather than left to defaults, and no logs exporter is set. See [Security & privacy](https://verity.md/docs/security).

**Retention.** The service deletes raw telemetry on a schedule: metric points after 14 days, spans after 30 days, and the raw batches as received after 7 days. It keeps the daily totals.

## Where to see it

- **Token telemetry**, for the open repository — what the agents in that repository spent: totals, trend, by model, by agent, by session, and the cost of each run on the run page.
- **Token telemetry by org**, across every repository you can read — grouped by the org that owns each remote, with a selectable window, and a link into each repository's own view.
- From the terminal, `verity status` summarizes recent activity and trend.

Costs are estimates

Dollar figures are imputed from token counts and published model prices, not authoritative billing. Use them to compare and trend spend, not to reconcile an invoice. The dashboard says so wherever a figure appears.

[Previous Org knowledge](https://verity.md/docs/org-knowledge) [Next The dashboard](https://verity.md/docs/dashboard)

Source: https://verity.md/docs/cost
