> ## Documentation Index
> Fetch the complete documentation index at: https://developers.reflection.ai/llms.txt
> Use this file to discover all available pages before exploring further.

# Overview

> Run evidence-driven security assessments with Mirror and Beam

Laser is an experimental harness for authorized penetration testing. It runs on [Mirror](https://developers.reflection.ai/mirror-overview.md), Reflection’s terminal coding agent, and uses [Beam](https://developers.reflection.ai/models.md) by default. It adds a shared evidence ledger, supervisor reviews, and workers for independent tasks. You can also configure an OpenAI-compatible model endpoint.

> [!NOTE]
>
> Get the source from [reflectionai/laser](https://github.com/reflectionai/laser). Repository access is currently required.

> [card] [Quickstart](https://developers.reflection.ai/laser-quickstart.md)
>
> Install Laser and run a local lab engagement.

> [card] [Developer guide](https://developers.reflection.ai/laser-developer-guide.md)
>
> Configure models, workers, Docker, and live controls.

## How Laser works

The `laser` launcher prepares a workspace and records the authorized scope. The **executor** investigates inside Mirror, saving observations, traffic, and findings to the **evidence ledger**. It can delegate independent tasks to **workers** sharing that ledger.

At review checkpoints, a separate **supervisor** model reads a brief and proposes the next plan. It cannot execute tools. Laser checks the plan against recorded progress before the executor continues.

```mermaid theme={"theme":{"light":"vitesse-light","dark":"vitesse-dark"}}
%%{init: {'theme': 'base', 'flowchart': {'diagramPadding': 24}, 'themeVariables': {'fontFamily': 'Reflection Inter, sans-serif', 'primaryColor': '#8a9a7b26', 'primaryBorderColor': '#6f7f61', 'primaryTextColor': '#2f3a27', 'textColor': '#6f7f61', 'titleColor': '#6f7f61', 'lineColor': '#6f7f61', 'edgeLabelBackground': '#f8f6f0', 'clusterBkg': '#8a9a7b12', 'clusterBorder': '#8a9a7b', 'actorBkg': '#8a9a7b26', 'actorBorder': '#6f7f61', 'actorTextColor': 'currentColor', 'actorLineColor': '#8a9a7b', 'signalColor': '#6f7f61', 'signalTextColor': 'currentColor', 'noteBkgColor': '#8a9a7b26', 'noteTextColor': 'currentColor', 'noteBorderColor': '#8a9a7b', 'labelBoxBkgColor': '#8a9a7b26', 'labelBoxBorderColor': '#6f7f61', 'labelTextColor': 'currentColor', 'loopTextColor': 'currentColor', 'git0': '#6f7f61', 'git1': '#a3875a', 'git2': '#5a7fa3', 'git3': '#a35a6f', 'gitBranchLabel0': '#ffffff', 'gitBranchLabel1': '#ffffff', 'gitBranchLabel2': '#ffffff', 'gitBranchLabel3': '#ffffff', 'commitLabelColor': 'currentColor', 'commitLabelBackground': '#8a9a7b26'}, 'themeCSS': 'margin: 0 auto; .edgeLabel .label, .edgeLabel span { color: inherit; } .dark & .edgeLabel, .dark & .edgeLabel p, .dark & .labelBkg { background-color: #1a1b1a; } .dark & .edgeLabel rect { fill: #1a1b1a; }'}}%%
flowchart TD
    E["Executor in Mirror"] -->|"Observations<br/>and results"| L["Evidence ledger"]
    L -->|"Progress brief"| S["Supervisor"]
    S -->|"Plan checked<br/>by Laser"| E
    E -->|"Independent tasks"| W["Workers in Mirror"]
    W -->|"Shared evidence"| L
    classDef default color:currentColor
```

Tool captures preserve evidence; findings and attack relationships must be recorded explicitly. Review the saved support before accepting a result: an agent’s `verified` verdict is not independent reproduction.

## Monitor and control engagements

Use `/laser` inside Mirror to expand the native panel, or `/laser-brief` to print its cached decision brief.

The [web UI](https://developers.reflection.ai/laser-developer-guide.md#web-ui) can launch engagements, monitor terminals, send follow-up input, and stop runs and their workers. It supports host execution and configured Docker profiles.

Overview summarizes outcomes and evidence issues. Evidence compares saved captures; Findings shows each claim’s support; Attack graph shows recorded relationships. Graph connections alone do not establish successful execution.

## Choose an engagement mode

| Mode | Use | Default objective |
| - | - | - |
| `ctf` | Controlled flag challenges | `flag` |
| `redteam` | Authorized assessments of running systems | `getshell` |
| `src` | Bug-bounty-style testing of live applications | `report` |

`src` is not a source-code auditing mode. The CLI defaults to `redteam`; the setup wizard offers `ctf`. Set a different objective with `--objective`. A supervisor stop does not necessarily mean success.

## Scope and execution boundaries

> [!WARNING]
>
> Test only authorized assets in an isolated environment. URL targets authorize the entire host, including all ports and paths. Laser’s HTTP scope checks do not sandbox shell commands or browser subrequests; apply independent network restrictions and use engagement-specific credentials.

Keep the web UI bound to loopback. Stopping an engagement does not contain detached host processes; see [shutdown and teardown](https://developers.reflection.ai/laser-developer-guide.md#stop-a-run).

## Related documentation

See the [developer guide](https://developers.reflection.ai/laser-developer-guide.md) for configuration and operational details, and the [Responsible Use Guide](https://developers.reflection.ai/responsible-use-guide.md) for Reflection’s usage guidance.
