New updates and improvements about V7 Go

GPT-6 Astra is available in conversations. A source reference now shows the exact quoted text behind an answer, and a run transcript keeps every user turn rather than a summary of them.

Models

• GPT-6 Astra: Available for use in conversations.

Answers and sources

• Sources show the quote: A source reference is attributed by the quoted text behind the answer rather than by a citation tag, so checking where a response came from is a matter of reading it.

• Complete run transcripts: Every user turn in a run is kept, so an audit or a playback shows the whole conversation instead of part of it.

• One search for agent skills: Agent Skill discovery runs through a single search across the platform, rather than a separate lookup per surface.

Clearer failures

• Temporary and permanent read failures read differently: When content cannot be read, the response says whether it is worth retrying.

• Still indexing is not the same as not searchable: Search and grep distinguish content that is still being prepared from content that cannot be searched, so an empty result explains itself.

Speed and reliability

• Background results arrive together, sooner: Results are delivered as soon as the final background task finishes, which shows most when several actions run in parallel.

• Steadier nested runs: Nested steps no longer reach for tool patterns that are unsupported at that depth, so longer runs complete more predictably.

Reporting

• Workflows reports drop backing projects: Projects that back a workflow are excluded from its report, so the figures describe the workflow.

• One last-activity time: A user last activity time reads the same wherever it appears.

• Report pages stop jumping: Space is reserved for the scrollbar, so content does not shift as a long report loads.

Interface

• Tool results render in chat: A message that carries results from a connected tool displays correctly rather than breaking the thread.

Slideshow Tools v2 is on for everyone. The property tool picker names agent tools as Agents, and Decide for me answers the question in front of it instead of stepping past it.

Heads up

• Gemini 2.5 Pro has been retired: It no longer appears as a selectable model, so nothing can be pointed at a model on its way out.

Tools

• Slideshow Tools v2, for everyone: The updated slideshow tools are on by default, with nothing to enable first.

• Agents are named in the property tool picker: Agent tools are labelled Agents, so they are easy to tell apart from the other tools a property can use.

• Skill mentions read as references: A skill mentioned in a message renders as a clear, clickable reference rather than raw text, so you can see which skill was invoked and open it.

• Pass part of an object into a skill: A skill argument can reference specific fields of an object, so one skill covers more cases without a variant for each shape.

• Refreshed Salesforce skills: The Salesforce Pipedream skills have been regenerated against current actions.

Chat and questions

• Decide for me decides: It answers the question in front of it rather than moving past it, so it completes the choice instead of skipping the work.

• Redesigned suggested options: Option cards are cleaner to scan, which makes a run of questions quicker to move through.

• Chats resume after a skill finishes: A successful skill run hands back to the conversation on its own, with nothing to restart.

• Notifications survive long work: Progress updates and handoffs are delivered durably, so a long run still reports back.

Workflows and reporting

• A dedicated card for a workflow run: A run opens in its own card, so progress and the next step sit in one place.

• Workspace usage reflects real limits: System limits are included, so remaining capacity reads against the boundary that actually applies.

• Internal tools are marked in usage reports: Internal tooling is named and flagged in Mission Control usage, so a line item does not have to be guessed at.

• Provider details read consistently: Provider information is presented the same way across reports.

• Snapshots keep short-lived users: A report snapshot carries temporary user details, so it matches what was true at the time.

Accessibility

• Project tables and inline fields announce properly: The project table grid structure and inline text fields carry the names and roles a screen reader needs.

Skills work in cases for everyone now, with no access request in the way. Reports name the guests and temporary grants behind your data, and a gate only runs again when something it depends on has actually changed.

Skills

• Skills in cases, for everyone: Skills are available in cases without special access, so a connected tool can be used directly in the case where the work is happening.

Workflows

• Gates rerun only on a real change: A gate re-runs when something it depends on moves, rather than on every pass, so a run does less work to reach the same result.

Reporting

• Guests and temporary access are visible: Reporting tracks temporary grants and identifies guest users, so who had access to what, and when, can be answered from the report.

• Column tooltips on the Users report: Hovering a column header explains what the metric counts, without leaving the page.

Integrations

• Older integration components keep working: Workflows built on earlier Pipedream component versions continue to run, and the Notion skills have been regenerated against the current SDK.

• Connected tools stop disappearing: A broad search no longer hides tools that are enabled on a connector, so an available tool stays available.

Access and interface

• Stay signed in through secure app access: A browser session survives, so reaching the product that way no longer signs you out repeatedly.

• Consistent initials on avatars: An avatar uses proper name initials rather than the first two letters, so people are easier to tell apart at a glance.

• The sidebar control sits where it belongs: The expand and collapse icon is no longer painted behind other elements.

Any reporting view can be taken away as a CSV, every view says how fresh its data is, and an email alias can now be verified from a link and used to mail into a case.

Reporting

• Views download as CSV: A reporting view can be downloaded as a CSV, so the same numbers can be worked on outside the product.

• Each view says how fresh it is: A view reports when its data was last refreshed, so a number can be read with its age attached.

• Active users count people: Workflow active users are counted from human actions that can be attributed to someone, rather than from everything that touched the workflow.

• Configured properties in workflow reports: Workflow reports show the properties actually configured, and property counts agree wherever they appear.

• Dimensions read as names: Reporting rows carry named dimensions, so a breakdown reads as names rather than identifiers.

• Series run in order: A reporting series is returned in chronological order.

• Faster reports: Labels are resolved once per dimension per slice and a region’s reads run concurrently, which cuts the time a report takes to return.

Accounts and access

• Email aliases are verified: An alias is checked during invitation and identity setup, a verification email is sent, and the alias is confirmed from the link in it.

• Mail into a case from an alias: Case chat integrations recognize an email alias, so mail sent from one reaches the case it belongs to.

• V7 people do not take a seat: A V7 user in your workspace is excluded from the seat limit.

Issues

• Act on many issues at once: Issues can be multi-selected and moved through their lifecycle together, including dismissing a batch.

• Active count on the list: The Issues list shows how many issues are still active.

• Reopened issues keep their history: An issue that has been reopened shows that history in the list.

Chat

• Generated files appear as tiles: Files a response produces are shown as tiles in the chat rather than named in the text.

• Consistent preview tiles: File previews in chat and in All data are capped as wrapping 4:3 tiles, so a wide file no longer distorts the row.

• Answered questions stay in place: An answered question panel renders at its position in the transcript instead of moving to the end.

• Context at a glance: A ring on the chat shows how much of the context window is in use.

Interface

• Readable subtle text: The lowest-contrast text tone has been raised and the tone below it removed.

• Accessible names restored: Counters and tree controls announce their names again.

• Filters keep their state: Filter state holds instead of resetting as you move around.

• Talk to an expert works: The "Talk to an expert" link points somewhere again.

Reliability

• Files that are not UTF-8 read correctly: Go Agent V2 reads plaintext that is not UTF-8 without failing on it.

• Steadier context assembly: Todo-state reads are guarded, so assembling context does not fail on a missing or malformed entry.

• Integrations stay put: Affected integration action components are pinned to a version, so an upstream change does not alter a step under you.

Reports can be read across a whole organization instead of one workspace at a time, output samples show on the instructions page with a mark when they have gone stale, and a field edited by hand now lands in the token report with the time it was made.

Reporting

• Reports across an organization: Requests carry an organization identity from end to end, so a report can be read for a whole organization, and each report says which organization produced it.

• Workspace on every row: Organization reports return the workspace behind each row, so an organization can be broken back down into the workspaces it is made of.

• Current month by default: Organization reports open on the current month rather than an arbitrary window.

• Manual edits reach token reports: A field edited by hand is recorded at the moment it is written, with a timestamp, and token reports read from it.

• The workflow name on the detail view: Opening a single workflow shows its name rather than its identifier.

• Cleaner Users and Chats views: Usage that resolves to no user is dropped from the Users view, and usage that resolves to no chat from the Chats view, so the series count only what they claim to.

• Field names say what they count: The user fields in the Workflows report are renamed to describe the thing being counted.

• Exact report scope: A report returns only what falls inside its scope.

• Chat reads are current: A request that reads chats waits on chat lifecycle freshness first, so a chat created or deleted a moment ago is reflected.

• One organization per report viewer: A report viewer can no longer end up assigned to more than one organization.

• Faster activity extraction: Activity fields are indexed by workspace and time, which cuts the time to extract them.

Workflows

• Output samples on the instructions page: The instructions page shows a preview of the output sample, so the shape of what a property returns can be checked in place.

• Stale samples are marked: A sample that no longer matches the instructions it was generated from is flagged as stale rather than shown as current.

Chat and agents

• Escape closes previews: Chat artifact and file preview overlays close with the Escape key.

• Background work finishes: A background sub-agent run stays alive past the turn that started it instead of being cut off with its parent.

• Agent Skill search heals itself: The state of the Agent Skill search index is visible and recovers on its own rather than staying silently broken.

• Slow tools stop hanging: A tool result that takes too long is bounded in Go Agent V2, so a slow tool fails rather than stalling the run.

• Vision rows no longer crash notifications: A replayed vision row no longer brings down Go Agent V2 tail notifications.

Interface

• Overflowing filters collapse: Filters past the width of the bar collapse behind a "+N" instead of pushing the layout.

• Status dots are announced: Status indicators carry an explicit accessible name, so a screen reader describes them consistently.

Reporting has a view for each thing you might ask about — workflows, chats, users, and models — and it now answers questions down to the individual chat. The controls you click all day share one refreshed design.

Reporting

• Views for workflows, chats, users, and models: Reports has a view for each, so consumption can be read against the thing you are asking about rather than one combined total.

• Workflow detail: Open a single workflow from the list and see its own report.

• Chat-level reporting: Reporting answers questions down to the individual chat, including its name, its message count, and whether it has been deleted.

• Budget snapshots in reports: Reports read budget snapshots, so spend can be read against a cap over time rather than only where it stands now.

• Names instead of user IDs: Top users and the Workflows view show reporting user names, and fall back to the email address where a name is missing.

• Only attributed usage in the Workflows view: Usage that cannot be attributed is dropped from the view, so the totals reconcile.

• More complete report data: Workflow reports return property types, reporting dimensions capture them, and chat rows carry their workspace so multi-workspace analytics separate correctly.

• Filter and count fixes: Metrics stay visible when you change the selected period, facet counts are exact across the full scope, measure filters resolve against the merged activity, and the month in the token report date picker is no longer off by one.

• Reporting access on your profile: Your profile shows the organization reporting access you have.

Interface

• One design pass across the controls: Switches, checkboxes, radio buttons, toggles, badges, tabs, and breadcrumbs share one visual treatment, and the workspace selector has been reworked with them.

• Keyboard and screen reader fixes: Menus are navigable by keyboard and announced correctly, and nested menu behaviour works again.

• The model is named: The Go Agent V2 settings tooltip names the model actually in use.

• Steadier rendering: Workflows render without the flicker on load, dates and times are handled the same way throughout, and action bars match the current design.

Reliability and speed

• Faster on wide projects: Adding an entity to a project carrying many properties is substantially faster.

• Transient database errors retry: The tool runner retries a transient database error rather than failing the field.

• Faster startup: Behaviours load without pulling in every module, which cuts startup time and memory.

• MCP routing scoped per message: Connector routing is enforced per message, preventing one connector interfering with another.

PDFs open faster, which shows most on the long ones. Resources can be downloaded or opened full screen from the viewer, and any email alias linked to your account will sign you in to it.

Documents

• Faster PDFs: PDFs load and render more quickly. The difference is largest on long documents.

• Resource viewer controls: Download a resource or open it full screen from the viewer.

• Hub files open on click: Click a row in a hub to open the file.

Access and interface

• Sign in with an email alias: Any email alias linked to your account signs you in to the right account, which helps where people carry more than one work address.

• Report filters as pills: Report filters use pill controls and region selection is a single choice.

• Consistent controls: Segmented controls, tabs, and navigation lists share one visual treatment across the product.

Reporting timeframes resolve in a timezone you choose, which matters once a team is spread across regions. Property groups stay with a workflow when you move it somewhere else, and sampling a collection is much faster than it was.

Reporting and cost

• Timezone-aware reports: Pick a timezone in the report filters and see timeframes resolved in it. Reports line up with local dates and business hours.

• More accurate chat costs: Chat cost is summed from actual token usage, so the number you review is the number you are charged for.

Workflows and collections

• Portable property groups: Property groups stay intact when you move or share a workflow between environments, so there is less to rebuild after an import.

• Faster collection previews: Sampling a collection loads much faster, which shows most on large ones.

• Stuck uploads fail cleanly: An upload that gets stuck fails clearly rather than hanging while the project waits on it.

Administration

• Delete unused organizations: Admins can remove an organization they no longer need without a support request.

• Workspace ownership stays visible: A mapped workspace still shows the organization that owns it.

Chat and documents

• File previews from chat: Click a file reference in chat to preview it in place, without losing the conversation.

• Expandable JSON lists: A truncated list expands from its "and N more" summary, and large previews stay capped at a readable size.

• Favorites in a collapsed sidebar: Favorites stay reachable when the sidebar is collapsed.

• Clearer model errors: When a provider rejects a model, the error names the model instead of leaving you to work it out.

Ask user question is on for everyone now, so a workflow can stop and collect input without any setup first. Documents open in readable JSON by default, and the citations inside them point where they should.

Workflows

• Questions available to everyone: Ask user question no longer needs enabling. Any workflow can pause and collect human input where it needs it.

Reading and evidence

• Readable JSON by default: Documents open in a readable JSON view, and citation links inside them render and resolve correctly.

• Citations in the right place: JSON output places citations on the fields they belong to instead of repeating them across nested keys.

• Grounding survives dotted field names: Source grounding stays intact when your data contains dotted field names, and searches referencing those fields run correctly.

Reporting

• Reporting respects permissions: Reporting analytics honour workspace and report permissions, including when a filter names a workspace outside your scope.

• Chats report summary: The Chats report opens with summary cards carrying the key numbers.

• Budgets scoped to the workspace: The overview budget card shows the budget for the workspace you are actually in.

Terra is the model workflows use now unless you pick another one. Source deletion finishes cleanly even when something interrupts it, and a bad property schema tells you what is wrong instead of failing quietly.

Models and extraction

• Terra by default: AI workflows use Terra unless you choose otherwise. You get the current model without changing any settings.

• More consistent date extraction: Dates pulled from documents and inputs resolve more consistently, so fewer extracted periods come back incomplete or wrongly normalised.

Sources and citations

• Reliable source deletion: Deleting a source completes cleanly, even if something interrupts it. You can retry safely without duplicate work or broken references left behind.

• Citations that land on the right file: A citation from a collection file opens the source it belongs to, in the right subproject.

• Grounding stays with the project: An uploaded input keeps its link to the right project even when its name is adjusted behind the scenes.

Properties and interface

• Property schema errors you can act on: An invalid property schema returns a clear error naming what is wrong.

• Number properties lose the New badge: The number property type is treated as a standard part of the workspace.

• Cleaner Mission Control cards: Metrics sit in more consistent cards, so the numbers are easier to read at a glance.

• Favorites sidebar: The favorites scrollbar lines up with the sidebar edge.

Case responses can now be rated, and a rating you regret can be changed. Model usage has moved into Mission Control, workspace reports keep their history, and there is a new icon set across the app.

Feedback on cases

• Case response ratings: Rate a case response directly in the product. There is an API for rating case messages too, so scores can feed whatever you already use to track quality.

• Ratings you can change: Change a rating after you have submitted it. A misclick or a second look no longer means the score is stuck.

Usage and reporting

• Model usage in Mission Control: Model usage appears in Mission Control, next to the rest of your workspace activity. One fewer page to open to answer which model is doing the work.

• Workspace report snapshots: Workspace reports keep historical snapshots of their dimension breakdowns. You can see how a number moved rather than only what it is today.

• Updated Luna and Terra pricing: Pricing for the OpenAI Luna and Terra models now reflects current rates, so cost estimates and usage reporting line up with what you are actually charged.

Interface

• A refreshed icon set: A bespoke duotone icon set runs across the app. The interface reads more consistently, and the common actions are easier to pick out.

• Chat connection dropdown fixed: The chat configuration page stays responsive when you open the connection dropdown. It used to freeze there.

You can now favorite any workflow or chat and get back to it in one click. Hit the star on a card to add it, and it appears in the Favorites section of the left navigation alongside everything else you have starred.

What they are useful for

• Recurring work: star the workflows you run on a cadence, like a weekly reconciliation or a monthly report, so you are not searching for the same one every time.

• Picking up where you left off: star a chat and return to that exact session later. The context is already there, so a similar task continues in the same thread instead of starting cold.

• Bookmarking a working setup: when a workflow is configured the way you want it, star it. It becomes the version you come back to and adapt, rather than one you rebuild.

Favorites are per user, so starring something does not change what your teammates see in their own navigation. Add and remove them as often as you like.

Workflow properties keep a version history now. There is an inspector to compare two versions and preview what a restore will do before you apply it, so rolling back is no longer a guess.

Versions and history

• Workflow property versions: Workflow properties keep a version history. You can see how a property changed over time instead of guessing which edit broke a run.

• Version inspector: Compare two versions side by side and preview what a restore will do before you apply it. Labels in the navigation rail are clearer, so you can tell which version you are looking at.

Reporting and calculations

• Hourly usage reporting: Usage data breaks down by the hour, and transformation progress appears in reporting. This is rolling out to selected organisations first, so it may not be switched on for you yet.

• Calculations that recover: A calculation that hits a database error resumes instead of failing. Long-running work continues rather than restarting from the top.

Stability

• Structured search stability: Searches that use JSON schema validation complete instead of crashing. You can keep narrowing a search without losing your place.

• Property changes apply immediately: A change to a property's JSON configuration is detected properly. No more cached configuration quietly overriding what you just set.

Interface

• Overview columns stay put: Your column choices in Overview survive a page reload. Set the table up once and come back to the same view.

• Refreshed date picker: The date picker matches the current design across the product. Scanning a calendar and choosing a date is a cleaner job than it was.

• Custom MCP connector badges: Status badges in the custom MCP connectors table size correctly, so the table scans evenly.

The last stretch of July leans towards the plumbing that keeps V7 Go connected to everything else. MCP connections can authenticate with an API key, images and PDFs now go through the same vision path, and nothing a run produces gets lost on the way out.

MCP and integrations

• API key authentication for MCP: MCP connections can authenticate with an API key. This helps where OAuth is impractical, such as a service account or a headless deployment.

• MCP OAuth discovery to spec: OAuth metadata is discovered in the order the MCP specification requires, so compliant services connect on the first attempt.

Documents and files

• One vision path for images and PDFs: Image crops and PDF pages run through the same vision processing. Results stay consistent across both, so you do not need to prepare a scan differently from a photo.

• Automatic case files: Files a run creates are saved as case files when the run finishes. Nothing has to be rescued by hand from a completed run.

• Artifact links become file cards: An artifact path containing a space or ending in punctuation renders as a file card instead of plain text.

• Clear out stuck library files: Delete library files that got stuck while processing. A blocked upload is now something you can clear yourself rather than a support ticket.

Budgets and reports

• Budgets in Mission Control: View and manage budgets from Mission Control, next to the work the spend belongs to. Budgets themselves shipped on 14 July; this is where they now live.

• Reports across workspaces: Mission Control reporting runs on the multi-workspace API, so a report can cover more than the workspace you happen to be in.

• Longer report date ranges: Mission Control reports accept extended date ranges. A quarter of activity fits in one report instead of three stitched together.

Sessions

• Cleaner session expiry: An expired session now ends in a consistent sign-in prompt. You get a way back in rather than an error screen from whichever feature noticed first.

Two new Gemini models, a better answer to "where did our tokens go", and one fewer reason to email support: organization admins can change their own settings now.

Models

• Gemini 3.6 Flash and Gemini 3.5 Flash Lite: Both are available for extraction and AI work. Flash Lite is the cheaper choice when the task is simple and the volume is high.

Cost and usage

• Token usage by dimension: Group token usage by the dimensions your team cares about instead of reading a single workspace total.

• Cost attribution for extraction and calculation: Usage detail shows which work created which cost, so an expensive extraction is traceable to the thing that ran it.

• Usage stats in Mission Control: Activity and key numbers appear in the Mission Control overview, rather than on a separate page.

• Reports that reflect reality: Reports show live deletion status and budget detail in the results, so you are not cross-checking against something else.

Administration

• Organization settings, self-serve: Organization admins manage their own settings without a support request. Workspace mappings and reporting access stay consistent across regions.

• Property history and restore: View a property's previous configurations and restore one when a change turns out to be a mistake.

• Email replies per workspace: Set how email replies behave for each workspace instead of applying one rule everywhere.

Fixes

• Citation highlights on scans: Citation highlights line up correctly on scanned PDFs and image-only documents, where there is no text layer to anchor to.

• Delete property dialog: The dialog shows its full warning and buttons. You can read what you are about to delete before you confirm it.

Reports have a dedicated overview, admins have finer control over who can see them, and runs now stop themselves before they get expensive.

Reporting

• Reports Overview: A dedicated section for reporting, so there is a place to start rather than a set of separate pages to hunt through.

• One report across workspaces: Generate a single consolidated report covering several workspaces in the same region, instead of pulling exports and joining them by hand.

• Sortable workflow reports: Sort, filter, and page through workflow detail reports. This matters once a report runs to thousands of rows.

• Report access at organization level: Admins can grant reporting visibility to specific people rather than opening it to everyone.

Cost and limits

• Run limits: A run stops before it runs too long or consumes too much. A misconfigured loop now costs you minutes.

• Chats stop at the billing limit: Once a workspace reaches its billing limit, chats stop. The limit is now actually a limit.

• Calculations that finish under load: Calculated fields complete when many updates land at once, instead of sticking half-done.

Reading and review

• PDF citations out of the way: Citation markers on a PDF no longer cover the text underneath.

• One source's markers at a time: Select a source and only that source's citation markers stay visible. Everything else steps back.

• Code blocks left alone: A code block in chat stays exactly as written. No more "Unknown item" pills appearing inside a snippet.