Frequently asked questions

What are WebArm24 Pipelines?

Pipelines are multi-step workflows: an ordered list of steps whose outputs feed the next step. You build them on a visual canvas (drag blocks, nest steps inside containers), configure task variables, test individual steps, and inspect run history. Pipelines suit chained automations — scrape or call an API, transform with JS or Python, branch on the result, notify Slack or email — without maintaining separate scripts for each hop.

What step types can I add?

Each step has a short name shown in the editor. Supported types:

• Agent — real-browser task with an LLM (navigate, extract, act on pages).
• HTTP — REST call with templated URL, headers, and body.
• Email — send via SMTP; optional attachments from sandbox output paths.
• JavaScript — sandboxed V8 script; input is the previous step’s output string; use return for the next step.
• Python — sandboxed subprocess (bubblewrap); input is pre-bound; use print() for output. No runtime pip install — use libraries baked into the server image.
• If / else — run one nested branch based on a condition.
• Parallel — run named lanes concurrently; merged output is JSON keyed by lane name.
• Loop — repeat nested steps by count, while, or until (with a max iteration cap).

Steps can be skipped with enabled: false without changing {{step_N_output}} numbering.

How do variables and outputs chain between steps?

Templates in URLs, bodies, and task text support:

• {{prev_output}} — output of the step immediately before.
• {{step_1_output}}, {{step_2_output}}, … — specific step (1-based index).
• {{variable_name}} — pipeline task variables (API keys, IDs) you fill before a run.
• {{user.id}} — when the previous output is JSON, dotted paths resolve like variables (task variables override same names).
• {{trigger_body}} — raw body of an inbound webhook POST (empty for GET); always set on webhook-triggered runs and not overridable by caller variables.

What is Build with AI?

In the Pipelines editor, Build with AI opens a chat panel backed by /api/pipelines/builder/chat. Describe the workflow in plain language; the assistant proposes steps with a live preview, sets the pipeline name and task variables, and can suggest cron, one-time run_at, or completion webhook settings. You edit the result on the canvas and save when ready. It is a drafting assistant, not a substitute for reviewing steps and secrets.

How do Loop, Parallel, and If / else work?

• If / else — conditions include text contains, regex match, JSON path equals, or agent decides (LLM true/false). Only then_steps or else_steps run.
• Parallel — each lane starts from the same {{prev_output}}; lanes run at the same time; the following step receives JSON like {"Lane A": "…", "Lane B": "…"}.
• Loop — count (fixed repeats), while (check before each iteration), or until (check after). Always bounded by max_iterations. Nested steps can include any step type, including other containers.

Right-click a nested block on the canvas for move, duplicate, copy, enable/disable, and remove — actions apply to that step, not the parent container.

How can I run or schedule a pipeline?

• Run now — button in the editor or POST /api/pipelines/{id}/run (JWT).
• Cron — 5-field expression in UTC (e.g. 0 9 * * * daily at 09:00 UTC).
• One-time — run_at Unix timestamp; cleared after it fires.
• Inbound trigger URL — unique /api/webhooks/pipeline/{id}/{token}; POST body becomes {{trigger_body}}. Optional inbound trigger only disables Run now except via trigger or schedule.
• Completion webhook — POST to your webhook_url when the run finishes; optional HMAC in X-Pipeline-Signature.

Import and export pipeline definitions as JSON from the sidebar for backup or sharing between accounts.

How do I use the Pipelines API?

Sign in for a JWT, then use /api/pipelines to create, update, list, delete, run pipelines, and read run history. Builder chat lives under /api/pipelines/builder. For one-off agent jobs without a full pipeline, use POST /api/agent/jobs or WebSocket /ws (documented on the API / Jobs pages).

What else does WebArm24 offer?

Beyond Pipelines: site crawling and per-domain knowledge cache, the Web Agent Playground for simpler scheduled automations, async HTTP agent jobs, OAuth/JWT accounts, and an optional embeddable widget. Compared to crawl-only tools (e.g. Firecrawl-class APIs), WebArm24 adds agent reasoning, multi-step workflows, and outbound integrations in one stack.

Which AI models and UI languages are supported?

OpenAI GPT-4o / GPT-4o-mini and Anthropic Claude (Haiku, Sonnet), configured on the server. Pipeline agent steps and Build with AI use the same stack. UI: English, Swedish, Norwegian, Danish, Finnish, German, French, Spanish, Dutch, and Polish.

Official URLs

Canonical: https://webarm24.online/ — Pipelines at /pipelines. Mirror: https://www.webarm24.online/. Use one base URL consistently for OAuth and webhook callbacks.