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

> ## Agent Instructions
> IronBee (ironbee.ai) is an AI QA engineer: it verifies code changes against the running app and keeps the evidence. It is not related to the IronBee open-source web application firewall.
> The console is at https://console.ironbee.ai.
> Install the CLI with `npm install -g @ironbee-ai/cli`, sign in with `ironbee login`, and set up a project with `ironbee install`.
> In GitHub workflows, use the action as `ironbee-ai/ironbee-action@v1` and pass the `IRONBEE_API_KEY` secret as the `ironbee_api_key` input.
> In GitLab pipelines, include the component `gitlab.com/ironbee-ai/ironbee-gitlab/verify@v1`; it reads the `IRONBEE_API_KEY` CI/CD variable by default.
> In Bitbucket Pipelines, run the pipe `docker://ghcr.io/ironbee-ai/ironbee-bitbucket:v1` and pass `IRONBEE_API_KEY: $IRONBEE_API_KEY` and `BITBUCKET_OAUTH_TOKEN: $BITBUCKET_OAUTH_TOKEN` under its `variables:`.
> Console scenarios (https://docs.ironbee.ai/console/scenarios) and CLI saved scenarios (https://docs.ironbee.ai/cli/guides/scenarios) are separate features.
> For plans and prices, link to https://ironbee.ai/pricing; the website is the source of truth for them.

# Platform and local verification in Bitbucket

> Where the Bitbucket Pipe runs the verification, how it picks from the repository's visibility, and which variables apply to each mode.

The IronBee Pipe can run the verification in two places:

* **Platform mode:** on IronBee. The console calls this **On IronBee**.
* **Local mode:** inside the pipe, on your Bitbucket Pipelines runner. The console calls this **On your Bitbucket Pipelines runner**.

Everything after the verification works the same way in both modes: the verdict, the fix, the commit and the report. Only a run on IronBee can be [published on the pull request](https://docs.ironbee.ai/bitbucket-pipe/guides/running-in-pipelines.md#the-report) by IronBee itself. The `VERIFICATION_MODE` variable picks the mode: `auto` (the default), `platform` or `local`.

***

## Compare the two modes

| | Platform (on IronBee) | Local (on your runner) |
| - | - | - |
| Where the agent runs | IronBee's infrastructure | Inside the pipe, in your pipeline |
| Model access | IronBee's | Your Anthropic API key or Claude Code OAuth token |
| Pipeline setup | None: no Chromium, no DevTools install | Chromium, IronBee DevTools and Claude Code, installed on every run |
| Anthropic credential | Only to fix what it finds | Required |
| What it verifies | Your web app, in a browser | The [DevTools platforms](https://docs.ironbee.ai/bitbucket-pipe/guides/verification-platforms.md) you enable: browser, backend, Node.js, Python, terminal |
| Evidence | The IronBee Console | The IronBee Console and the step's artifacts |
| Private repositories | Need the repository's workspace [connected to IronBee](https://docs.ironbee.ai/integrations/bitbucket.md) | Work with any checkout |

***

## How `auto` decides

With `VERIFICATION_MODE: auto`, the pipe decides from two things it can check without calling IronBee: whether the repository is private, and whether an Anthropic credential is passed.

| Repository | Anthropic credential | Mode |
| - | - | - |
| Public | Any | Platform. IronBee can check out a public repository without the Bitbucket connection |
| Private | Passed | Local. The pipe can't check whether the Bitbucket connection covers the repository, so it uses the engine it knows can run |
| Private | Not passed | Platform, the only engine available. This needs the repository's workspace connected to IronBee |

The first line of the plan in the step log names the mode and the reason.

> [!TIP]
>
> If your repository is private and its workspace is [connected to IronBee](https://docs.ironbee.ai/integrations/bitbucket.md), set `VERIFICATION_MODE: "platform"` explicitly. Otherwise `auto` runs on your runner whenever an Anthropic credential is passed.

A private repository whose workspace isn't connected can still be verified on IronBee: set `IRONBEE_BIND_REPOSITORY: "false"` to verify the running application without a checkout of the repository.

***

## Variables that apply to one mode

The pipe warns in the step log when you set a variable that doesn't apply to the mode it runs in.

**Platform mode only**

| Variable | Purpose |
| - | - |
| `IRONBEE_PROJECT` | The project the run reports to. Empty derives it from the git remote |
| `IRONBEE_ENVIRONMENT` | The project environment whose variables and secrets the run uses |
| `IRONBEE_BIND_REPOSITORY` | Bind the run to the repository so it verifies the changeset |
| `IRONBEE_JOB_TIMEOUT_MINUTES` | The run's timeout on IronBee |
| `APP_RESTART_COMMAND`, `APP_LOGS_COMMAND`, `APP_WAIT_SECONDS` | An app that runs in the pipe, reached through a tunnel |
| `APP_HEADERS`, `APP_SECRET_HEADERS` | Request headers for a deployed URL |

**Local mode only**

| Variable | Purpose |
| - | - |
| `IRONBEE_BROWSER_DEVTOOLS`, `IRONBEE_BACKEND_DEVTOOLS`, `IRONBEE_NODE_DEVTOOLS`, `IRONBEE_PYTHON_DEVTOOLS`, `IRONBEE_TERMINAL_DEVTOOLS` | The DevTools platforms the agent verifies through |
| `IRONBEE_EXTRA_CONFIG` | Raw IronBee CLI config. See [Custom configuration](https://docs.ironbee.ai/bitbucket-pipe/advanced/custom-configuration.md) |
| `CLAUDE_CODE_MODEL`, `CLAUDE_CODE_MAX_TURNS`, `CLAUDE_CODE_ARGS` | The Claude Code agent in the pipe |

`CLAUDE_CODE_CLI_VERSION` and the Anthropic credential also apply to the fix step in platform mode. See [Configuration](https://docs.ironbee.ai/bitbucket-pipe/configuration/configuration.md) for every variable.

***

## What's next?

> [card] [How it works](https://docs.ironbee.ai/bitbucket-pipe/concepts/how-it-works.md)
>
> The run from the plan to the report.

> [card] [Bitbucket](https://docs.ironbee.ai/integrations/bitbucket.md)
>
> Connect your Bitbucket workspace so IronBee can read private repositories.


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.