Cofactor / Changelog
◆ We run Cofactor on Cofactor · ship daily, recap weekly

The shipping
log.

Everything we've shipped, newest first. Plus a look at what's still cooking behind a flag.

§01 / Releases
NEWEST FIRST
2026-W31
July 27 – August 2, 2026
2026.W31
NewImprovedFixed

Platform hiccups now say so and retry smarter, and a page-aware AI assistant starts rolling out.

A big reliability week. When a run can’t reach a verdict because of a problem on our side, not your app, it’s now labeled "Platform error" everywhere instead of the ambiguous "inconclusive", and the platform does more of the retrying for you automatically. Assertions now verify at the exact point in the flow they describe, so a failure pinpoints the action that actually caused it. And in early access, a page-aware AI assistant is starting to roll out that knows what screen you’re on and can edit flows, debug runs, and answer questions right where you’re working.

  • newcliYou can now fork a workflow, or fork several at once with fork-map, straight from the CLI.
  • improvedrunsWhen a run can’t reach a verdict because of a problem on our side, an AI-provider hiccup or a platform issue, not your app, it’s now labeled "Platform error" everywhere instead of the ambiguous "inconclusive". The platform also retries smarter: after an immediate first retry, a second retry waits about a minute, giving transient outages time to clear.
  • improvedassertionsAssertions now run at the exact point in the flow they describe, not just at the end of a step, so a failure pinpoints the action that caused it and reports read cleaner.
  • improvedsettingsApplication settings now separate the everyday from the technical. General covers just the essentials, while session state, environment variables, and Release QA move to a new Advanced tab. Each section saves on its own.
  • improvedflowsYou can now re-enable a degraded flow right from the run dialog instead of digging through settings to find it.
  • improvedrunsRun videos are easier to follow: the cursor now highlights before an action happens instead of after.
  • improvedrunsRun details got a tidier header, and the technical labels for how a step ran now use plain terms, so it’s clearer at a glance whether a step replayed cleanly or needed extra help.
  • fixedtest environmentsA seed flow marked "run once per suite" no longer re-runs inside every test in that suite, which previously churned shared fixture data and made suites slower and less predictable.
  • fixedauthoringEditing an API sequence from a flow step once again opens the editor full-width over the flow canvas, instead of being squeezed into a narrow side panel.
  • fixedassertionsFixed a case where a test could be marked failed even though its assertion actually passed; the live verification is now always trusted over a stale pre-check.
  • fixedrunsFixed a handful of edge cases that could make a run stall or misfire on specific pages.
Sneak peekbehind a flag, rolling out to early teams
  • One assistant, everywhere: a page-aware chat that knows what screen you’re on, can edit flows, debug runs, and answer product questions, and remembers what matters to your team. Rolling out to early teams now.
  • Runs are starting to sign in instantly from a saved session instead of walking through the login form every time.
  • Something new is brewing for planning and approving what gets tested: early access rolling out now.
  • Something new for turning a written test idea into a real test: an early walkthrough that checks whether a proposed test is actually feasible before you commit to building it. Rolling out to early teams now.
  • Running your own test suites on Cofactor’s infrastructure now streams results live as they run, with video for each test. Available in early access, ask your Cofactor account manager for a look.
2026-W30
July 20 – 26, 2026
2026.W30
ImprovedFixed

The run detail page and browser-extension recording get more polish.

This week is about tightening things up. The run detail page got cleaner, recording a workflow with the browser extension got smoother, and file downloads recover more reliably mid-run.

  • improvedauthoringRecording a workflow with the browser extension got smoother.
  • improvedrunsThe run detail page header is cleaner and easier to scan.
  • fixedrunsFixed a couple of cases where a delayed or slow-starting file download could get lost mid-run.
Sneak peekbehind a flag, rolling out to early teams
  • Something new is brewing for teams with an existing test suite of their own: early access is rolling out now.
  • A strict, fast-fail run mode for pure regression checks is rolling out to early teams now.
2026-W29
July 13 – 19, 2026
2026.W29
NewImprovedFixed

Flows gain per-step deactivation, plus comments land in the Flow Designer.

Flows got more flexible this week. You can now deactivate any single step without deleting it, comment directly on steps in the Flow Designer, and file an issue straight from the CLI with the evidence already attached.

  • newflowsYou can now deactivate any step in a flow, from its menu in the Flow Designer or via the CLI, and runs simply skip it and continue. Deactivated steps show a clear tag, and a run is blocked with a helpful message if a later step still depends on data a deactivated step produces.
  • newauthoringYou can now leave comments on steps in the Flow Designer.
  • newapi sequencesFlows can now include a custom API step for calls the built-in step types don’t cover.
  • newissuesYou can now file an issue straight from the CLI, with the relevant code, logs, and video attached automatically.
  • improvedrunsTest runs now use a streamlined way of driving the browser that works directly with page structure, a simpler, more robust foundation for complex interfaces that trims navigation steps from minutes down to seconds on heavier pages.
  • improvedrunsThe AI that builds and runs your tests is now much better at pinpointing the exact control to act on in tricky interfaces, so steps on data-heavy pages succeed more often and recorded tests replay more reliably.
  • improvedrunsTests that trigger a file download now replay correctly from cache instead of needing to redo the download every run.
  • fixedai chatAI chat replies (Flow Planner, Flow Builder, run diagnostics, and the events explorer) no longer get cut off mid-answer. The full response is always saved, even if your connection drops while it’s streaming.
  • fixedrunsTest results and error messages are more accurate: an environment provisioning problem is now marked inconclusive instead of a false test failure, a credential-binding error now tells you what to do, and a couple of harmless background errors stopped surfacing as errors.
Sneak peekbehind a flag, rolling out to early teams
  • Something new is brewing for building tests out of reusable pieces, instead of repeating the same steps over and over. Rolling out to early teams now.
2026-W28
July 6 – 12, 2026
2026.W28
NewImprovedFixed

Seed data sets itself up automatically, and approvals no longer wait on a passing run.

Test environments just got a little more self-sufficient. Seed data can now set itself up automatically the moment an environment is ready, and the values it creates are available right inside your tests. Alongside that, approving a test no longer has to wait for a passing run, so your team can sign off on a test the moment it looks right.

  • newtest environmentsSeed steps can now run automatically the moment a test environment is provisioned, and whatever they create is available straight in your tests, no manual wiring required.
  • improvedflowsYou can now approve a test as soon as the spec looks right, no need to wait for a passing run first. Whether it runs in your suites still waits for a pass, so nothing unproven starts running.
  • improvednotificationsSlack notifications for a completed run now carry more useful detail at a glance.
  • improvedrunsThe Test Runs page got a round of polish: clearer empty states, clearer status labels, and long run titles that no longer overflow.
  • fixedrunsLonger suites no longer get cut off by an arbitrary time limit, and we fixed a bug that could mark a healthy suite run as stalled when it wasn’t.
  • fixedrunsRun timelines no longer show passwords or other secret values in plain text.
2026-W27
June 29 – July 5, 2026
2026.W27
NewImprovedFixed

Personas carry their own variables, the Prompt Library grows up, and your runs are watchable start to finish.

A big week for making Cofactor feel more like yours. Personas can now carry their own variables, so your tests reference the right account or token automatically instead of you wiring it up per environment. The Prompt Library grew into a real workspace you can browse, search, and tag. Tests that read PDFs and laid-out forms now line values up with the right field, and your runs are watchable end to end.

  • newpersonasPersonas can now carry variables that each credential fills in, so a test references the right account ID, tenant, or token automatically instead of you wiring it up per environment. Sensitive values stay masked.
  • newpromptsManage your prompts like a proper library. Browse and search from a side rail, read them full-size, and tag them so the right one is always a click away.
  • newtest environmentsSet up a test environment template's emulated integrations (Slack, Jira, Google sign-in, email) straight from the Cofactor CLI. The new email emulator even captures a live inbox, so sign-up and reset flows that send a code just work.
  • improvedrunsFast-mode run timelines now show every browser action a run performed, including ones from steps that were retried or failed partway, each described in plain language instead of a raw selector.
  • improveduiThe Link Repository and Edit Repository Settings dialogs got a roomier, cleaner layout that matches the rest of the app, with an easier repository picker and tidier monorepo settings.
  • improvedauthPassword sign-up and reset now accept long passphrases, so a memorable phrase works. When a password is not accepted, we tell you exactly why instead of a generic message.
  • improvedauthoringSmall comforts in the flow designer: the right panel is now resizable, and the Insert step menu is grouped, described, and wider so it is easier to find the step you want.
  • fixedformsTests that read PDFs (invoices, claim forms, EOBs, and other laid-out documents) now line values up with the right field instead of running them together, so assertions about what a form shows no longer fail by mistake.
  • fixedrunsThe Live Browser View no longer drops out partway through a run. It reconnects on its own after a hiccup, and momentary browser restarts no longer cut the stream, so you can watch a run start to finish.
  • fixedrecordingsRecordings are captured more faithfully, no more quietly dropped steps or rewritten values, so what you record is what you get.
  • fixedrunsRuns that recovered from a momentary internal hiccup are no longer mislabeled as flaky, so a "flaky" badge now means something you should actually look at.
Sneak peekbehind a flag, rolling out to early teams
  • Azure DevOps support is on the way. Connect your Azure DevOps project to bring work items and webhooks into Cofactor. Rolling out to early teams now.
  • Something's brewing to make your tests get smarter the more they run: the fast browser agent is learning to remember how each page of your app works and reuse that knowledge next time, so tests on familiar pages get quicker and steadier over time.
2026-W26
June 22 – 28, 2026
2026.W26
NewImprovedFixed

A new kind of AI review is brewing, plus recurring schedules and one-click fixes.

Something new is taking shape: an AI reviewer that walks your app the way a real person would and grades the experience, not just whether the test passed. Meanwhile, you can now put any run on a recurring schedule, fork a flow without losing a thing, and apply a suggested fix to a broken flow in a single click.

  • newschedulesPut any run on a schedule. Set a suite, a tag group, or an exploratory test to run on a recurring schedule using plain-English timing, and see what is coming up next in one place.
  • newrunsSee how big a run will be before you launch. The New Test Run dialog tells you exactly how many flows your selection is about to run.
  • newflowsFork a flow and keep its history. Forking now carries the original tags, inputs, scenarios, and config, stays linked to its source, and lets you suggest changes back, with a new lineage view of which flows depend on which.
  • newdiagnosticsOne-click fixes for broken flows. When a failure traces back to something fixable in how a flow is written, the diagnostic agent drafts the complete edit for you to review and apply in a single click.
  • newflowsRead your flow as a spec: flip any flow to a clean, copyable spec view with severity-colored assertions, and download it as a re-compilable file.
  • improvedexploratoryA smoother exploratory tester: it shows a plan it revises as it explores, records UX observations separately from real defects, asks a question instead of crying “bug” when something is ambiguous, and lets you re-run any test and browse its history.
  • improvedrunsClearer live run feedback: runs show up live without a refresh, with accurate live durations, suite tables that show the diagnosis instead of a raw error, and one consistent color for anything running.
  • improvedtest environmentsA faster, cleaner test environment: a rebuilt detail workspace, an agent inside the test environment that sets up and verifies an app from the inside, and an agent that warms up while your test environment provisions.
  • improvedauthoringAuthoring quality-of-life: environment variables in the “$” typeahead, agentic conditions that open to their instructions, a recorder that folds a whole form-fill into one step, and freshly generated flows that stay a draft until you run them.
  • improveduiA more cohesive look, with per-org colored icons and dialogs that match the rest of the app frame.
  • improvedtasksReopen a completed task to pick up follow-up work without losing its history.
  • fixedformsDate and date-time inputs now get the right value every time.
  • fixeddiagnosticsFailure diagnostics pull the screenshot from the exact moment of failure, and auto-diagnosis only fires on a real pass-to-fail regression, so you get fewer false alarms.
Sneak peekbehind a flag, rolling out to early teams
  • A brand-new kind of review: send an AI persona through your app and get back a graded read on the experience, with the rough spots ranked by severity and pinned to the run video.
  • A planner that reads your connected codebase, looking up real selectors, routes, and labels in your source instead of guessing.
2026-W25
June 15 – 21, 2026
2026.W25
NewImprovedFixed

Exploratory testing graduates to a feature of its own.

Our AI exploratory tester grew up this week: it moved out of Labs and into its own spot in the nav, so it is there whenever you want it. Point it at your logged-in app and it maps out real user flows, runs them, and hands back a report. Around that, runs got steadier under load and failure diagnosis got a lot sharper.

  • newexploratoryExploratory testing has its own home in the main nav. Aim it at a logged-in app and it discovers the journeys real users take, runs them, and reports back.
  • newschedulesSchedule your test suites: kick off a named suite from any automation, including on a recurring schedule, with an optional environment override. Suites open on their latest results.
  • newloginsLogins that fit your app: set a longer login timeout for multi-step verification, and runs use the environment tied to the login you are testing as.
  • newsupportNever miss a support reply. Get an in-app notification and an email that links back to the thread, plus an unread badge on the Support button.
  • improveddiagnosticsSharper failure diagnosis: Cofactor reasons over your connected source code to tell a real app bug from a flaky run, calls out mid-run browser crashes, and stays focused on what actually broke.
  • improvedflowsManage your flows faster: promote or pause a flow from its settings, filter the list by promotion state or health, and see the login persona clearly labeled.
  • improvedexploratoryTidier exploratory runs that open straight to their report, with pull-request runs showing the PR title in the list.
  • improvedrunsRun details, polished: tracked variables on the detail page, table rows that open in a new tab with ⌘/Ctrl-click, dialogs that keep their chrome as you scroll, and a degraded run that links back to its flow.
  • improvedflowsA flow whose single step is an API sequence now renders fully on the API tab.
  • fixedrunsSteadier runs: slow heavy pages are no longer mistaken for frozen ones, logins survive a mid-run recovery, a stuck step recovers instead of hanging, and one worker hiccup no longer takes the whole run down.
  • fixedloginsA passing login test now consistently shows success, and API tests keep their session fresh through logins that rotate cookies.
Sneak peekbehind a flag, rolling out to early teams
  • One exploratory run that logs in as several different people, so you can exercise role, permission, and tenancy behavior in a single pass, with results grouped by who was driving.
  • A new way to propose flow edits: stage your changes privately and submit them for review instead of touching the live draft, with the flow agent able to draft a change request for you.
2026-W24
June 8 – 12, 2026
2026.W24
NewImprovedFixed

A fresh coat of paint, and suites that skip the dead weight.

Big week. Cofactor got a head-to-toe refresh, your suites got smarter about not wasting their time, and you can now kick off a test without ever leaving your pull request.

  • newpull requestsRun tests straight from a pull request. Mention Cofactor and it posts back a single, tidy summary with sharper preview thumbnails.
  • newflowsSuites now spot a degraded flow and step around it instead of burning a whole run on a path it already knows is broken.
  • newinstallInstall the Cofactor skill for Codex straight from the CLI.
  • improvedbrandA fresh new look: redesigned docs, a refreshed homepage with a new personas section and the Venn brand mark, updated pricing, and dark mode across the app.
  • improvedreplaysSkim your replays at 2× and 4× in the player.
  • improvedflowsThe login step is now called “Acting As”, so it’s obvious whose session a test is walking through.
  • improvedflowsOpen a flow in a new tab with ⌘/Ctrl-click, and ad-hoc results split Plan and Report into their own tabs.
  • improvedflowsSet your own timeouts for the API calls that set up and clean up your test data.
  • fixedrunsCaptured values now stick around across every step, and runs recover more gracefully when something hiccups mid-flight.
  • fixedformsNative date pickers submit correctly, and resize steps do what you’d expect.
  • fixeduiFixed a Toast layout glitch and smoothed out a flicker on nav-link hover.
Sneak peekbehind a flag, rolling out to early teams
  • An easy, hosted way to wire Cofactor into your AI tools, with a one-click plugin and connector bundle.
  • A planner that scopes out an ad-hoc test before the browser agent dives in, so runs start with a game plan.
  • Cofactor is starting to really learn your app, with early work on deeper, code-aware product knowledge.