Workflows
Save a repeatable web-data task, run it, and retrieve verified results through your application.
Open Workflows in the console and choose a template. Save the source scope and settings as a named workflow. Every run freezes the selected revision. Results, files, and source evidence are available in the console, REST API, and MCP.
| Template | Inputs | Outputs |
|---|---|---|
| Structured Dataset Builder | Starting URLs and an extraction preset or field selectors | Validated records, CSV, JSON, JSONL, and field provenance |
| Knowledge Base Builder | Documentation URLs or saved documents, link depth, and chunk settings | Document manifest, source-linked chunks, and a ZIP bundle |
| Deep Research | A question, optional starting URLs and source constraints | Evidence-ranked report, exact source passages, coverage notes, and a ZIP with checksums |
| Competitive Intelligence | Official competitor sites and a comparison focus | Dated source observations, pricing units where stated, and changes from the previous complete run |
Deep Research ranks and quotes passages from collected sources, then checks each quote against saved evidence. Choose model assistance to group evidence from public pages on installations that support it. Reports still use verified quotations, and the JSON identifies the method. A source dossier is also available.
A partial run can contain usable evidence. Read quality.json alongside the result. Missing or inaccessible sources can appear as unavailable or unverified in a comparison. Each file in manifest.json includes an artifact ID, size, and SHA-256.
Dataset versions and refreshes
On the review screen, preview up to three records from the first starting page. Choose Suggest fields to find supported columns with observed values on that page. Add individual fields, then preview again before saving. The preview leaves the dataset version history untouched.
Choose an existing dataset in Structured Dataset Builder to save each usable result to its version history. You can also save a completed result to a new or existing dataset from its run page. Retries return the same version. The Datasets library keeps partial versions visible and uses the latest fully verified run as the comparison baseline.
Open a version to browse and filter records, inspect field evidence, and download the full CSV or JSONL export. Compare versions collected with the same sources and extraction settings. Missing observations appear as unverified.
Knowledge Base Builder reuses verified HTTP captures and their chunks when a compatible HTTP-only refresh confirms the source is unchanged. Refreshes still respect site limits and report unavailable sources. Open a document on the run page to browse its sections, read individual chunks, and download the original capture.
REST
Create an API key with workflows:run under API keys. Store it in LISTARO_API_KEY. The open-source docs page below is a live example.
export LISTARO_BASE_URL=https://listaro.vercel.app
curl -sS -X POST "$LISTARO_BASE_URL/api/v1/workflows" \
-H "Authorization: Bearer $LISTARO_API_KEY" \
-H "Content-Type: application/json" \
--data '{"name":"Python tutorial","kind":"knowledge_base","inputs":{"sourceUrls":["https://docs.python.org/3/tutorial/introduction.html"],"maxDepth":0,"renderMode":"never"}}'Use the saved workflow ID and revision from the response.
curl -sS -X POST "$LISTARO_BASE_URL/api/v1/workflows/$WORKFLOW_ID/runs" \
-H "Authorization: Bearer $LISTARO_API_KEY" \
-H "Content-Type: application/json" \
-H "Idempotency-Key: python-tutorial-2026-11-20" \
--data '{"revision":1}'Poll GET /api/v1/workflow-runs/{runId} until the run finishes. Use the same idempotency key and arguments when retrying an uncertain start. Use a fresh key for a new run. Download a file through GET /api/v1/workflow-runs/{runId}/deliverables/{artifactId}; its short-lived redirect grants access to that file. Follow the redirect without forwarding the API key.
A schedule pins a saved revision. Use the workflow's Schedule panel to enable, change, or pause it. The API returns the next eligible run time.
For Structured Dataset Builder, set inputs.existingDatasetId on a saved revision, or post {"name":"Reference data"} or {"datasetId":"<dataset UUID>"} to /api/v1/workflow-runs/{runId}/dataset. The run endpoint returns its dataset and version IDs in publishedDatasets. Use /records and /compare?baselineRunId=<run UUID> under the same run path to read rows and compare versions.
CLI and SDKs
Download the developer tools bundle and extract it. Install the CLI with npm install -g ./javascript. Install the TypeScript package with npm install ./javascript from your application, using the path to the extracted directory. Python 3.10 and later can install the included client with python -m pip install ./python.
listaro-workflows templates
listaro-workflows start "$WORKFLOW_ID" python-tutorial-2026-11-20 1
listaro-workflows wait "$RUN_ID"
listaro-workflows download "$RUN_ID" "$ARTIFACT_ID" tutorial.zipImport ListaroWorkflows from listaro-workflows in TypeScript or listaro_workflows in Python. Both clients support starting, polling, scheduling, event pagination, and checksum-verified downloads.
MCP
Connect an MCP client to https://listaro.vercel.app/mcp with OAuth or a workflow-scoped API key in the Authorization header. The OAuth consent screen lets you grant workflow access. See MCP integrations for the connection details.
Use workflow_v1_list_templates, then workflow_v1_save and workflow_v1_run. Retrieve the same run with workflow_v1_get_run and obtain a file link using workflow_v1_download. Existing operation-level tools remain available under their original names.
Connect a desktop client
In Claude Code, add the server with claude mcp add --transport http listaro https://listaro.vercel.app/mcp. Open /mcp and follow the OAuth sign-in.
In Cursor, open the MCP settings, add an HTTP server named listaro, and enter https://listaro.vercel.app/mcp. Use OAuth when Cursor prompts you to connect.
In Codex, run codex mcp add listaro --url https://listaro.vercel.app/mcp --bearer-token-env-var LISTARO_API_KEY with a workflow-scoped key in that environment variable. Use codex mcp list to check that the server was added.
Saved workflows also include a Get code menu with the exact revision in CLI, cURL, TypeScript, Python, and MCP examples.
Migrating a Firecrawl workflow
| Use case | Listaro starting point |
|---|---|
| Search and scrape several pages into a cited report | Deep Research, or the search_web and scrape_page operations |
| Extract records across pages | Structured Dataset Builder, or scrape_batch for a one-time capture |
| Crawl documentation for RAG | Knowledge Base Builder, or crawl_site for raw crawl results |
| Track an official pricing or release page | Competitive Intelligence for comparisons, or a saved monitor for one site |
Check each response's coverage and artifact fields when moving an integration. The endpoint names, metrics, and output schemas differ by service.