Changelog

We reviewed how Stride AI meters its work and fixed the places where credits went out for nothing new. The diagram editor's Explain and Scaffold drawers show the answer you already have instead of running again, and Regenerate asks for a fresh one. The command palette asks the model only when you click "Interpret with AI", not after every pause in typing, and "Open in AI Chat" fills the composer instead of billing the question twice. The UI designer holds credits for the real request instead of a flat rate: about 112 for a default run, not 400. A Modernize repository analysis now reads every line of the code it analyzes, where before it read about 14% of it, so an analysis can take longer and use more credits than before, and one that is still running is no longer billed at its maximum. The AI Story Writer's "Tone / Style" setting and Suggest Scope's sprint stories now reach the model.

When usage crosses 70%, 90% or 100% of a limit, the notice in the bell and the Inbox now says where you actually are: 166 of 200 AI credits reads 83%, not 70%. Limits that count things, such as projects, seats, webhooks and API keys, read as a count, for example 4 of 5 projects.

Text you send to Stride AI is no longer altered on the way. Code containing sequences such as $' or $& (found in shell scripts and regular expressions), and template syntax such as {{user.name}} from Angular or Handlebars, now reach the model exactly as written. The diagram editor's find and replace also inserts your replacement text as typed: "$$ cost" used to come out as "$ cost".

PRD Studio's "Max requirements" setting now defaults to 10 instead of 60, so a first run gives you a smaller backlog to review. It is a hard cap: requirements past it are trimmed in document order, and the Studio tells you how many. You can raise it as far as 150 in Configure, and projects that saved their own defaults keep them.

When the AI provider refuses a request, the credits held for it are released and nothing is billed. Requests that time out or fail with a server error are still billed at their estimate, as before. If a model stops being available, Stride now moves to another after the first failure instead of retrying it, and its default models moved off the ones the provider retired. Billing and usage notices now reach workspace owners and admins in the bell and the Inbox.

The "Payment received" email now shows the service period the charge covers. Renewals used to print the month before, and a first charge printed a zero-length window; the amounts charged were always correct. Receipts now go only to workspace owners, not to every admin as well.

The home dashboard is rebuilt as a single view. A briefing leads with the project's health and trend, the most important fact about it, one "Do first" link into the module that owns it, and the sprint's real burndown, with any other recommendations underneath. Below it, a Connections ledger tracks blocked work, tests, components, defects, process and decisions against their targets, and the sprint matrix sits beside a rail showing what changed since your last visit. The tabs and the role switcher are gone, so everyone sees the whole picture: velocity is on Plan → Reports and the sprint page, and activity is in the Inbox. Each section appears as soon as its data arrives and holds its place while it loads, so the page no longer jumps. A workspace that hasn't subscribed yet sees its sample project drawn as a connected graph.

The Projects list's Lead column is now Team: up to three faces and a count of everyone who can reach the project, with the people added to it listed first. The old column usually showed the same person on every row. Library cards and rows now show collaborators' profile photos instead of initials.

A member other than an owner or admin who loaded the sample project, or resumed a PRD draft on the welcome screen, could create a project they then couldn't open. Both paths now add the member to the project they create, and a project created from a resumed draft gets its board columns.

When checkout can't open, it now says why and what to do, for example that your team uses more seats than the chosen plan covers, or that only owners and admins can start a checkout, and it shows the reference it asks you to give support. Stripe's checkout page accepts promotion codes again, and a free-sprints code whose offer has expired no longer blocks a monthly purchase.

Four of the five automation triggers could be chosen and saved but never fired. All five now run when the change happens: an issue is created, assigned or changes status, a defect's severity changes, or a sprint is marked Completed, once for each unfinished issue. Status changes from GitHub, GitLab and Slack fire rules too, and the template that moves unfinished issues to the next sprint can now be saved. Rules you saved earlier start running now, so it is worth reviewing them. Bulk edits and imports don't fire rules.

Moving an epic or a sprint to the trash detaches its stories, and restoring it used to leave them detached. The trash now remembers which stories the delete detached, and restoring the epic or sprint reattaches them.

Stride now works out today the same way everywhere, from your own calendar. Late in the day in time zones behind UTC, a story due today could read "Overdue 1d" on the board while the releases screen showed 0 days left.

Printing a report or saving it as a PDF now includes all of it, on every screen that offers it, instead of one clipped screenful.

A share link stops working the moment its project goes to the trash, and the owner can still list and revoke it. A workspace whose subscription has lapsed can still change passwords, two-factor settings and connected sign-in accounts. Permanently deleting from the trash now needs the same permission everywhere, and the offline cache keeps each signed-in session's data apart and clears it when you sign out.

We walked the whole path a new team takes, from sign-up to first work, and fixed where it stalled. A founder on one seat no longer sees a red seat-limit warning: a full plan shows as information with "Manage seats", and only a real shortage warns. After paying you land as a paid customer even when the payment confirmation is slow, and if the sample project fails to load you're told and can retry. Checkout errors now appear in your language.

On a story, any acceptance criterion can now become a test case, and Stride shows when a criterion is already covered. The sprint page shows the sprint's test health, and Test Cases gains a template manager to list, create, edit, clone and delete test-case templates.

Related-items panels on stories, components, decision records, processes and test cases have a "Suggest links" button that uses Stride AI to propose links to related work in your project. One share dialog now lists, creates and revokes share links on everything that can be shared, and more kinds of work can be shared. Workspace tags can be renamed or deleted from the Library tag picker, and search now finds test runs, exploratory sessions and diagram components.

The process analysis page has a "Refresh insights" button, and the Optimize library can set several processes to Draft, Active or Archived at once.

Coding agents connected over MCP gain tools to create, list, delete and suggest cross-links, turn an acceptance criterion into a test case, read a sprint's test health, refresh process insights, read decision records, and ask whoami, which tells an agent its workspace, who it acts as and what its key can reach. Every existing tool call works as before.

A link to work in the trash is now hidden until the work is restored, instead of showing as "Unknown", and links are removed when either end is permanently deleted. Creating and removing links follows your role, archived projects and your plan everywhere links appear, and test runs now show their linked work.

On the board and in the backlog, a second edit to a card's due date, points or assignee no longer fails with a message that someone else edited it when you're working alone, and a change that saved is never reported as reverted. When a move is refused, the message now says why, for example that the column no longer exists. Cards can be dragged from anywhere on the card again, and date pickers show their icon in dark mode again.

Saves no longer fail with "Too many requests" because of activity you didn't start. Background work, such as command-palette suggestions and presence, now has its own allowance, and everyday limits are much higher. Bulk moves, bulk edits and reordering send one request instead of one per story, and selecting all past 200 stories now works. A story field that fails to save retries on its own, and repeated errors collapse into one message.

Editing a story's description, on its page or in the board's card drawer, could lose text when a save was slow or failed, and could blame another author who didn't exist. Saves now go out in order, a failed save keeps your text and tries again, changing another field no longer overwrites what you're typing, and you're told someone else changed the description only when someone did.

Cards issued in India need an active e-mandate for renewals, and a mandate can stop working weeks before the next charge. Stride now notices and emails the workspace's owners and admins right away, in their own language, pointing them to support, because a new card can't revive a mandate. The past-due banner also offers a direct way to contact support.

A coding agent using a wrong, revoked or expired key now gets a clear unauthorized response that says how to fix it, instead of an empty server error. Seven tools were added so agents can delete and restore diagrams, processes and designs, and list designs. In Stride AI, every tool now shows its proper name in all 12 languages, and cancelling a sprint, removing a project member or changing a member's role asks for your approval first.

Import is now a migration tool: export an in-flight project from Jira, Linear, Asana, ClickUp, Azure DevOps, monday.com, Trello or GitHub and carry on in Stride with its history, people, comments and structure. From Jira, epics arrive as real epics with their children, comments keep their authors and dates, created and completed dates are kept so velocity and cycle time work from day one, and issue links, fix versions and formatted descriptions carry over. A preview maps each person and shows everything the import will create. Imports of up to 20,000 rows are supported, large ones run in the background with progress, and one click undoes an import while keeping anything edited since.

Prototypes can now move on after a delay, scroll to a section and animate between screens. Each UI design can carry its own design tokens, with spacing and type-scale presets. For handoff there is a spec export and measure-on-hover, and designs gain share links, named versions with thumbnails, and select same or select inverse.

The UI design editor gains the direct manipulation you expect from a design tool: snapping while you resize, rotation, draggable corner radius, rulers with guides you can drag, a gradient editor, and pasting or dropping images. Styling goes deeper too, with object fit, aspect ratio, blend modes, text shadows, grid rows and spans, true centering, and rich text with its own color, size and weight on any selection, set in six embedded open-license font families.

Every diagram type Stride draws (ERD, BPMN, system context, container, component, deployment, sequence, data flow, state machine) is now readable by external coding agents: get_diagram returns the semantic graph with resolved connection names, and list_diagrams discovers what a project has. link_diagram binds a diagram to a story, defect, or epic, and get_story_context now serves linked diagrams and process models alongside designs, docs, and tests, so an agent implementing a ticket sees the architecture it must build against.

An agent key used to be bound to exactly one project, so working across two meant minting two keys. When you generate one you now choose what it can reach: a fixed list of projects, or every project you can reach — which follows your access as it changes. A fixed list is the default, and the agent is refused everywhere outside it, including projects you can open yourself. Inside the projects it can reach, a key is still capped by your own role there.

Revoking an agent key left its integration in place, which then blocked re-creating one for the same project — the connect form showed a database error, and the row itself offered no Rotate or Revoke button to recover with. Both are fixed: every agent row now carries Rotate and Revoke whatever its state, and generating a key for a project you previously revoked simply works.

Cursor joins Claude Code and Codex as a first-class Stride agent runtime. Settings → Integrations → Cursor mints a scoped agent key and offers a one-click "Add to Cursor" install, or the config to paste into ~/.cursor/mcp.json yourself. Every workspace member gets their own key, and all three runtimes share the same MCP endpoint and the same tool surface. Agent-authored work is now easier to spot everywhere it appears: each runtime writes under its own name and brand mark, so a comment, an activity row or a story's Reporter field shows "Cursor agent" with the Cursor logo rather than a generic avatar. Run history moved to a provider-neutral URL (Settings → Integrations → your agent → Executions); old links redirect.

Design and Optimize are no longer Pro-only — Starter now ships all four modules, and grows from 3 to 5 seats, 3 to 5 projects, and 150 to 200 AI credits per seat. Pro moves to 1,000 AI credits per seat. Starter is $15/seat/mo and Pro is $39/seat/mo; annual billing is now a flat "2 months free" on both. Enterprise publishes a $69/seat floor and adds a private-cloud deployment option. Existing Starter and Pro customers keep their current price for 12 months and get the new, larger limits immediately.

Pro's per-seat storage allowance grows from 2 GB to 5 GB, raising the workspace ceiling from 200 GB to 500 GB at the 100-seat cap. The increase applies immediately to every existing Pro workspace — no action needed.

Legacy Intelligence is now Modernize, rebuilt end to end. Analysis runs as a live staged pipeline you can watch, resume, and cancel; uploads accept the languages the feature was built for (COBOL, VB6, RPG, Fortran, PL/SQL, Delphi, ABAP, JCL and more); screenshots are truly analyzed with AI vision; chat streams markdown and remembers the conversation; and extracted requirements can become epics, stories, acceptance criteria, and tests in the same project with one click.

Modernize goes deeper than a single upload: point it at a whole repository and it ingests the codebase at scale, cites the exact files behind every finding in a dedicated Evidence tab, and captures tribal knowledge through a structured SME review and interview. A behavioral-parity gate scores how closely a proposed rebuild matches the legacy system, a portfolio view rolls cost, risk, and economics up across every system you track, and findings export to Word or straight into your Library.

A best-in-class sweep touched every module: clearer empty states, honest loading and error states, tightened spacing and typography, keyboard and screen-reader fixes, and dozens of small interactions that used to feel half-built. Roughly 97 individual improvements across Plan, Design, Optimize, and Verify landed in one coordinated pass.

Quality gates are now tunable per project: set your own thresholds for coverage, pass rate, and open-defect count, and the gate reads them instead of a fixed rule. Defects gained SLA tracking with due-by targets and breach flags, and test suites got first-class management so you can organize, reorder, and reassign cases without leaving the module.

Releases got a ground-up redesign with a clearer timeline and scope view; the project page opens on a health hero that reads status, risk, and momentum at a glance; and the calendar now lets you create work in place by clicking a day instead of opening a separate dialog.

Epics can carry a cover image and inline image uploads in their description, the same way stories do. Every Create dialog across the app now discards its draft cleanly when you cancel, so a half-typed story or defect never lingers to surprise you the next time you open it.

Sprints are no longer create-once: rename a sprint, shift its start or end date, and rewrite its goal right from the sprint view. The board, calendar, and reports all pick up the change immediately.

Deleting a story, defect, sprint, or most other items now happens instantly and drops an Undo into the corner. Change your mind and one click restores it, so destructive actions stop feeling risky.

Bring a backlog in from any tool: map each CSV column to a Stride field, preview the rows before anything lands, skip duplicates on re-import, and keep due dates on the right day regardless of the exporter’s timezone. Sprint and status values reconcile against what already exists.

The path from sign-up to a working, paid workspace got rebuilt end to end: a clearer plan and cadence selector, seat picking that matches your team, and a unified checkout that returns you to the right place whether you complete or cancel. New workspaces reach their first real screen faster.

The whole authenticated product is available in twelve languages: English, Spanish, French, German, Italian, Portuguese, Hindi, Mandarin, Japanese, Korean, Russian, and Arabic. Arabic renders fully right-to-left, numbers and dates localize correctly, and you choose your language in Settings.

You now get notified the moment a story is assigned to you, someone replies in a thread you follow, or a sprint you are on starts, and due-soon reminders arrive grouped instead of one-by-one. The weekly digest deep-links straight to the work, surfaces missed mentions, and skips quiet weeks, and a monthly what’s-new email recaps what shipped.

The Pro plan now scales to 100 seats and 200 GB of workspace storage at the same per-seat price, so growing teams can stay on Pro far longer before needing a conversation about Enterprise.

Open a story and ask Stride to draft the architecture diagram or the business process it implies. The result is a clean, aligned diagram you can edit on the canvas, linked back to the story it came from, so design work starts from real context instead of a blank page.

A dedicated analytics surface reads your real delivery data: hero KPIs, a signals engine that flags risks by rule, cumulative-flow and WIP-age charts, a quality two-by-two, team workload, and predictability. Every panel drills down to the exact stories behind the number, and the views export.

Every UI design you generate lands in a fast, filterable gallery with real server-rendered thumbnails, so a project’s screens are browsable at a glance instead of hidden one dialog deep.

From the backlog, hand a story, or a whole selection, to your connected coding agent (Claude Code or Codex) in one move, and pull finished work back through the inbox. The hand-off is connection-aware, so it only offers agents you have actually linked.

Generated UI designs can be played as clickable prototypes, so you can walk a flow instead of staring at a static mockup. A Fast tier returns a design in seconds while a Quality tier takes longer for a more finished result.

Two new MCP tools let an external agent load everything about a story: its description, acceptance criteria, comments, linked designs and tests, and related work. It can then mark the story in progress and begin, so the ticket becomes the single source of truth the agent codes against.

Sketch directly on the diagram canvas with a stylus or trackpad: continuous freehand ink that does not snap to the grid, for annotating an architecture diagram or roughing out an idea before you formalize it into shapes.

Describe a screen and Stride generates a real, styled UI mockup you can edit, refine element by element, and attach to the story it belongs to. Generation picks the right archetype for what you asked for and always returns a usable design, with a deterministic fallback when a provider is degraded.

Connect GitLab over OAuth and link merge requests and pipelines to your stories the same way the GitHub App works: a ticket key in an MR title or branch auto-links it, pipeline status flows back, and signed webhooks keep everything current.

The whole integrations surface got an honest, consistent rebuild: clearer connect flows and status for GitHub, Slack, Jira, GitLab, and agents, a redesigned API keys and reference area, and one unified activity feed for API calls, webhook deliveries, and MCP tool runs.

The diagram editor reached best-in-class: entity-relationship diagrams, AI-assisted editing, a distraction-free focus mode, grouping, mixed content (sticky notes, text, tables, images, frames), Mermaid and SQL import, and export or share of any diagram.

The BPMN Process Designer gained the editing muscle you expect from a dedicated tool: a richer palette and quick-connect, a step inspector for duration, role, cost, SLA, and automation, live validation, and cleaner layout so imported and mined processes render legibly.

AI usage is now metered in clear credits instead of an opaque dollar cap. See your balance at a glance, buy credit packs when you need more, and opt into auto-top-up so long runs never stall. Every AI action shows what it will cost before you run it.

The billing page now shows the card on file, your next invoice and its amount, and countdowns for trial, grace, and cancellation windows, so there are no surprises about what you will be charged or when.

Import an event log and Optimize mines the real directly-follows graph: nodes heat by throughput, edges weight by frequency, the happy path and rework loops stand out, and per-step durations come from measured data rather than estimates. It is process mining on your actual history.

Chat and Operate merged into a single always-on Stride AI that answers questions, drafts and edits stories, tests, and PRDs, and traces work across Plan, Design, Optimize, and Verify. It reads and writes through the full tool suite, remembers your preferences, cites its sources, and refuses to invent ticket keys or numbers.

Talk to Stride AI: tap the mic in the composer and dictate your prompt with the browser’s speech recognition, hands-free.

Sign up or sign in with any Microsoft account: work or school (Microsoft 365) and personal (Outlook, Hotmail, Live). One click on the login or sign-up page; your Microsoft identity links to a Stride account with the same email, alongside Google and email/password. Two-factor auth and everything else work exactly the same.

PRD generation moved into a live orchestration screen: watch each stage stream in, keep a history of every generation, and re-run with a side-by-side diff so you can see exactly what changed. An apply funnel turns the finished PRD into a backlog in one pass.

The Dependencies page became an interactive graph: pan and zoom the web of blocks and blocked-by links, spot the critical chains, and jump straight to any item, instead of reading a flat list.

All nine Create dialogs (story, defect, test case, sprint, epic, release, ADR, process, and project) got a consistent pass: smarter defaults, inline AI assists where they help, and universal fixes so creating anything feels the same and fails gracefully.

Verify shipped a 17-PR module-wide overhaul: streaming AI test generation that preserves multi-step tests through the pipeline, drawer-first test-case UX with traceability, AI-assisted defect analytics, and per-workspace AI budget provider locking. AI test cases now link back to stories cleanly in the dashboard.

Plan sprints V2/V3 land as a Jira-equivalent module: analytics, scope-creep detection, sprint planning views, and Jira-style structured release notes.

Rebuilt /projects from the ground up: single-lead column model, slimmer filters, saved-view tabs you can pin per workspace, quick-action hover menus, scannable sprint + delivery chips on every row, and auto-complete when a sprint closes. The PAUSED status got dropped; "in progress" is now binary.

Roadmap got a premium final-form pass: contextual zoom on hover, an intelligence panel that surfaces dependencies + risk + capacity at a glance, status filters, search, density-toggle, and full keyboard navigation. The Gantt now reads like Linear’s project timeline.

Kanban cards got a ground-up redesign: due-date chip, child-item count, strict color economy so only the things that matter (overdue, blocked) light up, and overflow rules so long titles don’t break the column rhythm.

Workspace owners can now grant per-project access (read / write / admin) instead of all-or-nothing workspace membership. Enforcement runs at every layer (API routes, MCP tools, and every UI navigation guard) so a contributor with read-only access on Project A and admin on Project B sees and writes exactly that, everywhere.

Activity feed and Inbox merged into one sidebar surface with sections, snooze, and a "done" pile so everything that wants your attention (assignments, mentions, sprint changes, status updates) lives in one place with the same triage flow.

Stories now carry a real dueDate from schema through to UI: sidebar field on story detail, bulk-edit modal on the backlog, sync both directions with Jira’s due-date field. Kanban cards and the Plan calendar already pick it up.

New Plan Calendar: month / week / day views with event cards, drag-and-drop to reschedule due-dated work, an Unscheduled panel for triage, sprint bands and release pins overlaid on the canvas, and keyboard-accessible drag.

Multi-select on the backlog now means business: group-level select-all checkboxes, a bulk-edit modal that updates labels / assignee / sprint / status / due-date on every selected story at once, and bulk drag-and-drop that moves whole selections between sprints or backlog sections in one motion.

The MCP server gained 25+ new write tools: sprints (create / update / transition), epics, story relations, comments, attachments, labels, bulk updates, restore / delete operations. Claude Code and Codex can now run a sprint planning session start-to-finish on your behalf.

Plan Automation Phase 1 ships: pick from a curated template library, edit any rule inline, test-fire a rule against a real story before turning it on, and audit every execution in a history pane. The whole flow got a redesign so building "when X, do Y" rules feels less like Zapier and more like part of Plan.

Six-PR Optimize module landing: Stride Process Stream + Process Explorer header, Variants + Conformance tabs in Analysis, live insight generators from event streams, a Monte Carlo simulation engine wired to a server-backed run API, contextual Insight Strip surfaces on Plan story detail, and cross-module Process Stories narratives.

Drop a screenshot, log, or attachment directly into a story description or comment: paste, drag-and-drop, or click-to-upload all work, with inline previews for images and a download chip for everything else.

Every story now carries a stable Jira-style key (e.g. PROJ-123) derived from its project’s prefix and auto-incremented per project. Keys appear in URLs, breadcrumbs, search, Slack / Jira / GitHub mentions, and the universal Cmd+K palette.

Process Designer ships as a full BPMN 2.0 canvas: rich palette (tasks, events, gateways, swimlanes), step inspector with field-level editing, version history with diff, autosave, import / export to standard BPMN XML, and a generate-handoff button that turns the diagram into a runbook.

Generate Backlog from meeting notes got rebuilt as a multi-step streaming wizard: paste notes, AI extracts initiatives, you accept / reject / edit each before any story lands. SSE-streamed so cards appear as the model writes them.

Every comment now supports emoji reactions: pick from the standard set or open the full picker, see counts and reactor avatars inline. Reactions sync over Inbox notifications so a thumbs-up doesn’t fire a fresh email.

Legacy Intelligence moves from prototype to first-class surface: list and detail routes for legacy systems, full CRUD, a shared data layer that the rest of Optimize reads from, and the "Legacy Intelligence" flagship positioning on the tour.

Workspace owners get a live AI usage dashboard: every AI call attributed to a user, feature, story, and model, with cost breakdowns, item-count badges, and a paginated log that powers chargeback or just keeps the spend cap honest.

PRD wizard ships its overhaul: two-phase generation (outline → stories) that parallelises across the outline, SSE-streamed progress per stage, per-story test batching, retry on per-stage errors, and a fallback heuristic writer that still ships real content when the AI provider is degraded.

Sprint detail pages gained inline add / remove story controls: multi-select from the backlog into the active sprint, pull stories out with one click, all without leaving the sprint view. Matches Jira’s sprint-management muscle memory.

Every AI feature that targets Anthropic now uses prompt caching for the shared-context portion of the call. Typical workspace prompts cache the system + workspace-summary blocks, cutting per-call cost dramatically. Updated cost-pricing table tracks the new floor.

Story descriptions, comments, ADRs, meeting notes, and PRD bodies share one Tiptap-based rich editor: same toolbar, same slash menu, same markdown / paste-from-anywhere, same keyboard shortcuts. Phase 3 starts wiring AI inline commands.

Stride now ships as a Model Context Protocol server: connect Claude Code or Codex to your workspace and let an external AI agent read your backlog, draft stories, fan out test cases, or open a PR with full project scope. Per-user OAuth, scope-aware tile UI, and project-RBAC integration so each integration only sees what its owner can.

Cmd+K is now the universal entry-point: search every entity, jump to any surface, run natural-language commands ("close PROJ-12, move PROJ-13 to next sprint"). Click any link inline to slide-in a peek drawer with stacked navigation so you don’t lose context. Central keymap provider with chord support.

New agentic AI chat lives in the app shell: answers product questions, drafts tickets and tests, traces work across modules. Three-phase rollout (foundation → tools → memory) shipped with migration 0065.

Install the Stride GitHub App on your org; any PR with a Stride ticket key in the title or branch auto-links to the story. Opt-in: PR-merge auto-transitions the linked story’s status. HMAC-signed webhooks, App-level auth (no per-user OAuth needed).

Slack v1 ships: OAuth install with workspace-scope, per-project channel picker, and configurable activity fan-out (stories, sprints, releases, mentions). Threaded so a noisy sprint doesn’t flood #engineering.

Google as an OAuth provider on both /login and /register. Workspace invites also auto-link a Google identity so an invited user lands in the right workspace on first sign-in.

New workspaces see a 5-step activation checklist on /dashboard: create your first story, invite a teammate, connect an integration, generate a PRD, configure billing. Each step deep-links to the right surface and persists progress.

An ambient right-rail follows you across the app: context-aware suggestions for the surface you’re on, prompt-hash-cached so identical context doesn’t re-spend AI budget. Smart cards surface the next-best action ("close 3 stale stories", "the sprint ends in 2 days, want a release-note draft?").

Any filtered list view (backlog, board, roadmap, projects) can be saved as a named view, shared via URL (filters / sort / grouping all encoded), and pinned to your dashboard as a live widget.

Create scoped API keys from Settings → API keys, then call the REST API with Authorization: Bearer nl_…. Full reference lives in Settings → API reference once you sign in.

Every activity now fans out to your webhook endpoints with an HMAC-SHA256 signature, exponential-backoff retries, a per-hook delivery log, and one-click replay.

PRD generation, meeting notes, and full data exports now run as background jobs with live progress so they don’t time out on large workspaces.

New /legal/security trust page, workspace data export as JSON, owner-only immediate workspace deletion, full audit log for OWNER/ADMIN roles.

Email verification, password reset, per-email login lockout (5/15min), constant-time bcrypt comparison, and audit log entries for every auth event.

Stripe checkout + billing portal, usage metering, and per-workspace monthly AI spend caps with 50/80/100% alerts.

Stride is now production-ready with Sentry error tracking, distributed rate limiting, legal pages, and a first-class deployment pipeline.