Product
Changelog
Full Progon changelog — all releases, newest first.
- 1.5.0
Version 1.5.0 redesigns the Progon workspace and unifies controls across the web app and extension. Session fixes preserve sign-in during temporary failures and coordinate refresh across tabs.
Changed- The web interface has a new design: navigation sits in a capsule in the header, with project, environment, and selected AI model controls nearby.
- Projects, scenarios, bugs, and the map have glass surfaces and a landing-style background.
- Buttons, fields, tabs, menus, and dialogs share their appearance and behavior across the web app and extension.
- Sign-in and registration screens have been redesigned, with dark, light, and system themes consistent with the rest of the interface.
- Redesigned screens distinguish loading, empty lists, and errors, with a retry action after a failure.
Fixed- Network failures and temporary server errors no longer end your session: sign-out only follows an explicit authorization rejection.
- Session refresh is coordinated across tabs, and requests are never replayed under a different session.
- 1.4.25
Add a screenshot or document to a message so AI can work with the source material. File support for the selected model is checked before sending.
Added- Attach PNG, JPEG, and WebP images, PDFs, plain text, and Markdown files to project chat messages.
- Drag files into chat or paste images from the clipboard; file tiles show attachments before sending.
- Chat explains in advance if the selected model does not support an attachment.
Fixed- Attachments are retained for retry when sending fails.
- 1.4.24
Give AI work in plain language and choose the allowed actions. The task produces a report of completed tests and findings.
Added- Describe application exploration, test preparation, or scenario execution to AI in plain text.
- Choose the work scope and permitted changes before launching.
- The report shows results, findings, and limitations of the completed work.
- Create bugs from findings after confirmation.
- 1.4.23
A swarm distributes work among browser agents coordinated by a lead. Launch history brings together progress, results, and requests for user input.
Added- A lead AI agent coordinates several browser agents during exploration and automation.
- Choose the agent count and effort for a launch.
- Inspect agent activity, tasks, and events, and stop the swarm when needed.
Changed- Shared launch history shows who started the work, where it runs, its outcome, and whether your response is needed.
- 1.4.22
Communicate with a running agent through chat. Questions, answers, and additional messages help exploration or automation continue without a new launch.
Added- A running agent sends its question to the conversation with an answer card.
- Answer through the card or assistant, and the agent continues with the response.
- Send additional messages to a running agent during exploration and automation.
- 1.4.21
The map brings together the application interface, API, and data. Captured traffic links to screens and actions and helps prepare requests for repeat testing.
Added- Switch between General, UI, API, and Data views on the application map.
- Inspect network requests linked to exploration screens and actions.
- Reuse API requests collected from captured traffic.
- Inspect data schemas, errors, and network request timings.
Changed- Passwords, tokens, cookies, email addresses, and phone numbers are masked in captured API traffic.
- 1.4.20
Web and extension notifications share a feed and read state. Updated extension panels help track steps and expected results.
Added- A personal feed and bell collect notifications about work and AI questions.
- Open the relevant launch, question, or conversation from a notification.
Changed- Read state and notification group preferences are shared between the web app and extension.
- Extension panels have updated element, checklist, and result cards, with step navigation during playback.
- 1.4.19
Step and checklist generation follows consistent rules across interfaces. Results distinguish created tests from description fragments that could not be converted.
Added- Generation results show text fragments that did not become steps.
Changed- Generate several concrete actions from a goal instead of a single vague step.
- Generation preserves preconditions, expected results, and the language of the description.
- Checklists account for the actual form fields being tested.
Fixed- If you cancel generation before saving begins, new steps are not added to the scenario.
- 1.4.18
Exploration, automation, and step recovery use consistent browser execution. Navigation and additional site windows can be handled without losing work, and stops have an explicit reason.
Changed- Windows opened by the site remain available for continued work, including sign-in through another site.
- When element recovery or step execution is impossible, the result shows a specific stop reason.
Fixed- Navigating to another page no longer destroys the running sequence of AI actions.
- A scenario can execute its final sign-out action in the application under test.
- 1.4.17
The assistant handles test work as a sequence of actions. It can wait for run results, continue after confirmation, and analyze completed work.
Added- The assistant can wait for a run to finish and analyze its result.
Changed- The AI assistant carries out multiple project actions in sequence within one conversation.
- After an answer or confirmation, the assistant continues with the saved context.
- The completion summary is grounded in saved action results.
- 1.4.16
Each application has an address for the selected environment. Scenarios use that address at launch, so recordings are not tied to a single host.
Added- Set addresses for each application and environment pair.
Changed- Create a project by name and add the application address later.
- A scenario opens at the selected environment address and uses its variables and login data.
- The extension identifies the application and environment from the site address, while recorded steps do not hardcode the host.
- 1.4.15
Automation reuses recorded navigation between pages. Shared steps expand into an execution sequence while retaining their relationship to the source scenario.
Changed- A recorded navigation step opens the saved address without a new AI decision.
- Shared steps expand into a sequence of actions during scenario automation.
Fixed- Executed action results retain their relationship to the source scenario or shared step.
- 1.4.14
Automation queues are stored on the server and survive closing the tab. Add scenarios to a running queue and return to its progress later.
Added- Add scenarios to an already running queue.
Changed- Make Auto queues keep their state on the server rather than in a single browser tab.
- Automation requests through MCP use the same queues and execution states.
Fixed- Return to the queue and continue tracking work after closing the tab or restarting the browser.
- 1.4.13
Questions and confirmations awaiting a response stay beside the message input. The conversation displays prose and action results in the order they arrive.
Changed- AI questions and confirmations stay above the message input and remain accessible in long conversations.
- Confirmation cards show readable project, machine, and environment names.
Fixed- Prose and result cards appear in arrival order; focusing the input no longer shifts a long conversation.
- 1.4.12
The AI assistant asks clarifying questions and helps choose a machine for long-running work. Job cards retain their status after a reload.
Added- Answer assistant questions using suggested options or your own text.
- The assistant discovers a suitable local runner and helps with connection instructions.
- Long-running work appears in a status card that survives reloading the conversation.
- 1.4.11
Assign a local runner to automate a group of scenarios or explore an application. The batch page shows progress and provides launch controls.
Added- Start Make Auto for a group of scenarios on a selected local runner from the web app.
- Send AI application exploration to the same connected machine.
- Track a batch, open live execution, and cancel the work.
- Machine enrollment uses a connection approval flow, with expired or rejected requests shown explicitly.
- 1.4.10
Manage external AI tool connections from Progon and confirm changes proposed through MCP. Reports show test stability and problem areas in the project.
Added- Manage MCP connections and personal access tokens with selected access scopes.
- Use MCP to discover connected runners, inspect a run, and request cancellation.
- Prepare scenario imports through MCP: confirm, reject, or undo changes.
- Reports include a project overview, test stability, weak spots, and bug hotspots.
- Open filtered bugs from a report and discuss results in chat.
- 1.4.9
A connected local machine runs scenarios in a dedicated browser. Dispatch a single scenario or a group from the web app and inspect results in Progon.
Added- Connect a local runner with a dedicated browser and the Progon extension.
- Launch a scenario from the web app and view progress and results.
- Select multiple scenarios for a batch launch on the connected machine.
- Installation instructions help you prepare and connect a local machine.
- 1.4.8
AI explores the application and collects knowledge for the project map. History and search help you revisit discovered screens and relationships.
Added- Launch AI application exploration to collect information about screens and actions.
- View exploration results and application map history.
- Read application zone summaries and search discovered knowledge.
- 1.4.7
Get started with Progon through a conversation and configure traffic capture with site rules. Calibration status shows what is already known about application network requests.
Added- Conversational onboarding helps you navigate to the relevant project sections.
- Configure site traffic capture rules; a recording session uses the selected rule version.
- The extension shows calibration status and observed website network request patterns.
- 1.4.6
Step generation follows the source description more closely: it preserves the language and expected results, with scenario length guided by the test content.
Changed- AI creates steps in the language of the test description.
- Removed the arbitrary generated-step limit so detailed descriptions are not truncated to a short scenario.
Fixed- Expected results remain attached to actions when generating steps from text.
- 1.4.5
Automation of a single scenario and a group uses a shared progress panel. User instructions are passed to AI with the launch.
Added- Collapse the progress panel while continuing to work on the page.
- Instructions supplied at launch are passed to AI for scenario automation.
Changed- Single and batch automation display progress in a shared panel.
- 1.4.4
Imports explain what happened to each group of tests. Skipped and rejected entries are shown separately from created entries.
Changed- Scenario and checklist import outcomes are displayed separately.
Fixed- Imports show separate totals for created, skipped, and rejected entries.
- Unnamed rows no longer disappear without explanation: they are included in the import outcome.
- 1.4.3
AI proposals in chat appear as changes you can review, confirm, and undo. Scenario links take you from discussion to the test.
Added- Chat shows proposed project changes in cards before applying them.
- Confirm changes from the card and undo operations that have been applied.
Changed- Scenario links in assistant responses open the relevant test.
- 1.4.2
Use AI to turn scenarios into automated tests and recover element bindings during playback. Captured API requests and test plans make prepared checks reusable.
Added- Make Auto converts written scenarios into automated tests and shows progress for each step.
- Save API requests from captured traffic for reuse.
- Organize scenarios into test plans and track their execution.
Changed- If an element binding is lost, AI can find the element again and resume playback.
- Variables extracted during a run survive navigation between pages.
- 1.4.1
External AI assistants can access Progon scenarios, coverage, and results through MCP. Discuss tests with the project data at hand.
Added- Connect an external AI assistant to Progon through MCP to read project data.
- Search scenarios and inspect their steps in execution order from an AI tool.
- Retrieve coverage recommendations and the latest run results through MCP.
- 1.4.0
Version 1.4.0 is the main AI capabilities release. For the first time, Progon lets you verify test results without writing code: just describe what should be on the page, and AI takes care of the rest.
Added- AI result assertions: describe the expected outcome in plain language, AI compares it against the real page after playback.
- CDP API traffic capture: network requests from a recording session are automatically turned into API steps in the scenario.
- AI checklist generation: one click creates a ready-to-use checklist for any scenario.
Changed- AI assistant panel redesigned: quick access to generation, assertions, and checklists from one place.
Fixed- Fixed side panel getting stuck on tab reload with an active scenario.
- Steps targeting elements inside Shadow DOM now replay correctly in fully automatic mode.
- 1.3.0
Version 1.3.0 focuses on manual testing: Progon now assists not only those who want full automation, but also those who prefer to perform steps themselves, delegating only result detection to the system.
Added- Manual playback mode: the user performs actions, the system detects and records the result in real time.
- Custom fields for scenarios: add arbitrary metadata — status, priority, assignee — and filter by them.
Changed- Selector self-healing reworked: fewer false positives on pages with dynamic class names.
Fixed- The Run button in a test run is now immediately active after the previous run completes.
Removed- Removed the legacy Selenium XML export (replacement: the built-in API export).