Skip to content

The cross-harness, self-improving memory layer for AI agents. Join our community: https://discord.com/invite/zdNChS2fBu

1.9 KB
Raw Blame
---
title: "beacon ci start"
description: "Start a detached Beacon CI telemetry session"
---

## Command overview

`beacon ci start` starts a local collector, writes a session state file, and prints or exports telemetry environment variables for later workflow steps.

```bash title="Command syntax"
beacon ci start
```

In the GitHub Action, these variables are written to `$GITHUB_ENV` so later `run:` or `uses:` steps can inherit them. Claude Code uses environment variables; Codex uses a generated `CODEX_HOME` with Beacon OTLP configuration.

## Flags

| Flag | Description |
|------|-------------|
| `--harness <harnesses>` | Harnesses to configure. Defaults to `claude,codex` when omitted |
| `--log-path <path>` | CI runtime JSONL log path |
| `--state-path <path>` | CI session state path |
| `--github-env <path>` | Optional GitHub Actions environment file to append exports to |
| `--forward <destination>` | Optionally forward CI events to a customer-managed SIEM. Supported values are `splunk` and `falcon` |
| `--forward-endpoint <url>` | SIEM HEC endpoint URL for the selected `--forward` provider |

## Examples

Start a detached Beacon CI telemetry session:

```bash title="Start a detached Beacon CI telemetry session"
beacon ci start --harness claude,codex
```

Start a session with a stable runtime log path:

```bash title="Start with a stable runtime log path"
beacon ci start \
  --harness codex \
  --log-path "$RUNNER_TEMP/beacon/runtime.jsonl"
```

## Related

<Columns cols={2}>
  <Card title="beacon ci" icon="terminal" href="/cli/ci">
    Review CI artifact paths and command group behavior.
  </Card>
  <Card title="beacon ci finish" icon="stop" href="/cli/ci-finish">
    Stop, validate, and export a detached telemetry session.
  </Card>
  <Card title="CI Telemetry Exports" icon="file-export" href="/log-forwarding/ci-telemetry-exports">
    Export CI runtime JSONL through workflow artifacts, S3, GCS, or downstream pipelines.
  </Card>
</Columns>