Skip to main content
Free plan, no cardStart free

FLOWQA workflows · updated 2026-11-10

Keyword-driven testing tool: QA workflows on a visual canvas

Keyword nodes on a canvas, run in a real browser on demand, on a schedule or from a webhook; 81 keywords, RPA steps, connectors and an exception desk.

For: QA engineers and automation leads who want tests and small automations built from named steps, operations teams that run the same browser job on a timetable, and anyone who needs a failed step to land on a desk with a due time rather than in a log.

acme-shop / checkout-flow.recIllustrative
1 selector healed#pay-now-btn[data-test="pay"]

Capabilities

A canvas of keyword nodes

Around eighty node types in sixteen families: browser, navigation and control, data, API, assertions, control flow, RPA, integrations, orchestration, authentication and research. Connect them, configure each one, save the workflow and run it; a recording can also run as a self-healing bot.

Read more: A canvas of keyword nodes

81 keywords, each schema-checked

Every keyword declares its arguments and the runner refuses a step that does not match. The catalogue covers frames, tabs and pop-ups, searchable dropdowns, date pickers, multi-page tables with a keyed diff, localised numbers and dates, saved sessions and a human input step that waits for a person.

Failures classified, not just logged

Every failure is classified from a table (retryable, needs a person, action), only idempotent steps are retried, compensation runs in reverse, and each failed item becomes an exception you can retry, hold, assign, dismiss or reopen, with a per-day trend and failures grouped by pattern.

Evidence per run

A durable row per item, a screenshot on failure, a Playwright trace per failed item, per-node results on the result page with the files a node produced, and a signed run manifest shown with the item.

RPA and connectors with your own credentials

Capture a page as a structured record, fill a form from it, verify it and branch on a decision table. Connector nodes for Slack, Telegram, WordPress, Reddit, Google Sheets and web search call the real service with credentials from your vault, and verify the account before they run.

Search & Collect

A research job as a workflow: search, collect the records, resolve each one’s licence, read scanned PDFs by optical character recognition, and deliver DOCX and XLSX files with the run and by e-mail. Free and Solo Pro get a taste of it: three results, no OCR, five minutes, a few runs a month.

A test-file library per user

Upload the files your workflows read, such as data sheets, documents and fixtures, into a library with a per-user allowance: 25 MiB on Free and Solo Pro, 50 MiB on Team and 100 MiB on Enterprise, up to 500 files and 45 MiB per file. Usage is counted from the files themselves, never from an estimate.

Schedules, webhooks, revisions, n8n import

Run on demand, on a schedule or from a secret-gated webhook. A workflow keeps revisions you can restore and promote between environments, and an exported n8n workflow can be imported as a starting point, with unsupported node types reported and embedded secrets stripped, sixty nodes at most.

Read more: Schedules, webhooks, revisions, n8n import

How it works

  1. Build the workflowDrag nodes from the palette, connect them, configure each one and save. Keyword Steps gives any of the 81 keywords the run’s page and variables.
  2. Run it against a real browserA run drives a pooled browser and records a result per node, with the page the step left behind on failure.
  3. Put it on a timetableA schedule, a secret-gated webhook, or both; each item gets a priority class and a due time.
  4. Work the exceptionsFailed items wait on the desk with their class, screenshot and trace: retry, hold, assign, dismiss or reopen, and read the failure patterns.
  5. Read the resultThe result page lists every node, the files it produced and the signed manifest of the run.

What you receive, and what we do not do

Evidence ledger
Per run
A result per node, the run history, a downloadable trace and a signed run manifest.
Per item
A durable row with its status, attempt and due time, a screenshot on failure and a Playwright trace.
Per exception
The failure class, the error, who holds it, its retry, dismiss and reopen history and whether its due time was missed.
Per Search & Collect job
Each record with its source, its licence and where the licence was read, plus the DOCX and XLSX files produced.

Limits we state

  • QA workflows are a Team feature. On Free the single tool gives 20 workflow runs a month, Solo Pro takes it as an add-on, and the 14-day Solo Pro trial does not include it.
  • The AI prompt node runs on a local model and its output is labelled generated and unverified; under load it can be refused by the concurrency cap, and it makes no external AI call unless your team buys and switches on Work with AI for the node.
  • A node that writes to an external system is a high-risk step: it is never run on its own by an automated planner, and outbound traffic goes only to allow-listed hosts.
  • A human input step waits for a person. If nobody answers by its due time, the item fails and opens an exception; it never passes by default.
  • Search & Collect reads a scanned document within the job’s time budget; a long scan can be read in part, and the record says how much was read.

Plans: QA workflow runs: 800 a month on Team and per contract on Enterprise; on Free as the QA workflows single tool, $15 a month for 20 runs; on Solo Pro as an add-on. Search & Collect in full on Team and Enterprise, with a free taste on Free (5 a month) and Solo Pro (10). See the metering table.

Questions buyers ask

What is keyword-driven testing?

A way of writing tests as sequences of named actions with arguments, such as open a page, type into a field, read a table, compare two values, so that a tester composes a test without writing code and the engine supplies the implementation of each keyword. Here the keywords are nodes on a canvas and entries in an 81-keyword catalogue, each with a declared argument schema.

Is this robotic process automation?

In part. The RPA nodes capture a page as a record, fill a form from a record, verify a page against one and branch on a decision table, and a recording can run as a self-healing bot on a schedule. It is built for testing and for small browser jobs, not as a replacement for an enterprise RPA platform.

Can I import my n8n workflows?

Yes, as a starting point: an exported n8n workflow is imported with unsupported node types reported and embedded secrets stripped, up to sixty nodes. You then replace what did not map and add credentials from the vault.

What happens when a step fails at three in the morning?

The failure is classified from a table. A retryable step is retried only if it is idempotent, compensation runs in reverse where it is defined, and the item lands on the exception desk with its screenshot and trace, a due time and, where set, an escalation. A human input step that nobody answers fails at its due time and opens an exception.

Do my credentials leave Dokimore?

Connector nodes use credentials from your vault, encrypted at rest and referenced by name; each connector verifies the account before it runs, and outbound calls go only to allow-listed hosts. Dokimore holds no shared integration account on your behalf.

Try it on your own application

The Free plan asks for no card and runs for 14 days after the trial. Or bring one flow from your product and we record it live.

  • Free plan, no card
  • 14-day Solo Pro trial
  • Demo replies within one business day
  • Data hosted in India