Health scores now reflect every problem signal in a user's traces, and each signal's rate can be read against the unit that fits it.
ImprovedThe Users health score now scores each trace for frustration and other negative emotions, contradicted claims, laziness, problems, and failed tool calls, then averages across the user's traces, so one bad trace no longer zeroes a user's score.
ImprovedThe health score distribution on the Users page is now a histogram.
AddedEach Signals tab has a rate card whose denominator you choose inline, such as tool calls, turns, tasks, or traces. Home signal pins use the same saved denominator.
Fixedmoda search now honors --conversation-id, finds names and non-English text, returns results for recent time windows, and no longer returns the same trace for every query.
Frustrated Users on Home
Home now tracks how many of your active users are frustrated, replacing a signal rate that sat near 100% for most workspaces.
ImprovedThe Signal rate tile on Home is replaced by Frustrated users: the share of active users whose health score is in the Frustrated band, with the change from the previous three days and a sparkline.
Website and Brand Refresh
moda.dev has a new look, and the brand page has more colors and logo formats.
ImprovedRestyled the landing and marketing pages with a new homepage hero, smaller type on phones and tablets, and a slimmer mobile navigation bar.
AddedThe brand page now includes amber, blue, sky, and teal, plus PNG and JPG logo downloads.
ImprovedReplaced the remaining legacy Moda logos and favicons with the current brand mark.
Full-Length Content
Long messages, tool outputs, and problem text are kept in full everywhere Moda shows or sends them.
FixedLong messages and tool outputs are no longer shortened when stored or analyzed.
FixedProblem descriptions, Slack and Linear messages, and list API responses no longer cut off long text.
FixedMessages written by your harness are no longer attributed to the user or the agent in problem findings.
Improvedmoda feedback accepts misroute and harness categories, and keeps unrecognized categories instead of rejecting the note.
Traces Described by Their Tasks
Each trace is now described by the tasks the user asked for, each with its own status, instead of a single AI-written summary.
ImprovedThe trace list and trace pages show each trace's tasks in place of a summary, with open tasks first.
RemovedRetired trace summaries. Previously generated summaries remain as history.
ImprovedProblems, and the traces inside each problem, are ranked by weighted impact rather than by affected users alone.
FixedTimes on problem cards are shown in your own time zone.
Problems Page and Configurable Home
A dedicated Problems page, a side panel for problem traces, a Home you can arrange yourself, and a Signals overview that shows each number once.
AddedNew Problems page with KPI cards, affected-trace and signal-rate charts, category cards with size-over-time timelines, search, filters, and a paged table of problems. It replaces Analytics in the sidebar.
AddedOpening a trace from a problem shows it in a side panel with what went wrong, the cited messages, false-positive review, Open in agent, and a link to the full trace.
ImprovedHome is rebuilt around key numbers, pinned signals and problems with charts, open problems that need attention, and your top use cases. Customize lets you show, hide, pin, and reorder sections.
ImprovedThe Signals overview shows one card per detector with its rate, change, trend, and top pattern.
ImprovedDashboard pages now open on the last three days instead of 7 or 30 days.
ImprovedEmotion and tool-failure detections on a problem appear as finding cards, and tool-failure cards point to the failed call.
ImprovedTool results in the trace view are shown with the tool call that produced them.
Signup Page and Safer CLI Writes
A stable signup URL, and CLI write commands that refuse to guess.
AddedAdded a /signup page for creating a Moda account.
ImprovedCLI write commands reject unknown flags and unsupported --dry-run instead of writing anyway, and are never retried automatically.
ImprovedExpensive CLI commands such as ask now have cost limits.
FixedThe Data API frustrations endpoint now reads the same emotion data as the emotions endpoint and the dashboard.
FixedCorrected status page ingestion checks and removed past false alerts from its history.
Readable Problem Evidence
Problem findings read like the conversation they came from, and traces render the same way for every source.
ImprovedProblem finding cards lead with a summary and list their cited messages in time order, labeled by who said them: the user, the agent, or a tool.
AddedFinding cards show a severity breakdown and mark the user's own call-out.
ImprovedOpening a finding's trace starts at the beginning of the relevant turn and highlights only that finding's evidence.
FixedDetector descriptions no longer appear as if the user had said them.
ImprovedTrace rendering is consistent across sources: tool calls, reasoning, harness context, and standalone results start collapsed, and long bodies share one expandable preview.
ImprovedUpdated the dashboard and site to patched Next.js and dependency releases.
Open in Your Coding Agent
Hand a trace, problem, custom signal, or use case to Cursor, Claude Code, or Codex with a prompt already written.
AddedAn Open in button on traces, problems, signals, and use cases opens Cursor, Claude Code (terminal or VS Code), or Codex with a prompt that links the finding and lists the moda commands that load its details. Nothing is sent until you press Enter.
AddedChoose a default coding agent and an optional GitHub repository in Settings, Integrations, Coding agents. Copy prompt works with any other tool.
Account and Organization Settings
Manage your profile, organization, and membership without contacting support.
AddedEdit your profile name, profile photo, theme, and time zone, change your password, and change your email with a confirmation link.
AddedA new Organization page lets you edit the organization name, slug, and logo.
AddedChoose which organization Moda opens when you sign in.
AddedMembers can leave an organization, owners can delete one, and you can delete your own account. These actions are grouped in a Danger zone.
ImprovedOnly the organization owner can demote or remove admins.
ImprovedThe activity log now records invitations and joins, and shows who made each change.
FixedRenaming an organization updates the organization switcher immediately.
Alert Destinations
Send one alert rule to several places and control where alerts can post.
AddedOne alert rule can now send to several Slack channels, Linear teams, and an email list, with a delivery result for each destination.
AddedChoose which Slack channels alerts may post to, and disconnect Slack from the Alerts channels tab.
Linear Fix Checks for Custom Signals
Link a custom signal to a Linear issue and let Moda check whether the fix worked.
AddedCreate a Linear issue from a custom signal, with its description, conditions, and recent hits, or link an existing issue.
AddedWith completion checks on, Moda compares the signal's rate before and after the linked issue is completed, waits at least seven days, and shows the result in the signal panel.
AddedA custom signal can alert several Slack channels, chosen from channels the Moda bot is already in.
AddedThe Data API emotions and frustrations endpoints can be scoped by user_id and conversation_id.
ImprovedCoding agents can do from the CLI what the dashboard does, including user lookups, undoing problem actions, and managing custom signals, with guards against silent wrong answers such as unknown profiles or ignored flags.
ImprovedThe moda.dev navigation shows Open dashboard when you are signed in, and the sign-in page skips the form when a session already exists.
FixedTrace rows that were delivered more than once now appear once on the trace page.
Alerts and the Custom Signal Builder
Alerts get their own page with email delivery, and every team can build custom signals from a list of conditions.
AddedA new Alerts page brings Rules, Channels, and History together, so you can see what is configured and what was sent. It is available to every workspace.
AddedEmail is now an alert destination for reports and threshold alerts. Recipients must be members of your workspace.
AddedThe custom signal builder uses a list of conditions, such as use case, tool call, and user emotion, that must all match. Drafts save as you work, and Find matching traces checks the whole signal against past traces.
AddedDescribe a rule in words and Moda writes the matching expression, or refine an AI judge's instructions in place.
ImprovedProblem traces show the quoted evidence and the reason it was included, name the signal that raised them, and open the cited messages in place.
ImprovedRedesigned the Modesius panel's starting state with suggested questions.
Linear Integration and Longer Sessions
Connect Linear in one click, and stay signed in reliably.
AddedConnect a Linear workspace, file a problem as a Linear issue or link an existing one, and keep its status in sync.
AddedLinear is available as an alert destination, so new problems can open issues automatically.
Addedmoda problem-linear and moda problem-linear-file, plus matching MCP tools, show and file Linear issues for problems.
AddedSlack is available on the Integrations page for every workspace.
AddedA What's new window in the dashboard highlights the latest weekly update after each release.
ImprovedBrowser sessions last 30 days and survive temporary backend errors with an automatic retry instead of sending you back to sign-in.
AddedAn optional inactivity timeout of 15, 30, or 60 minutes in Settings, Account. It is off by default.
ImprovedThe status page now publishes measured availability from checks that run every five minutes, with 90 days of history.
RemovedThe Hallucinations tab is no longer shown on the Signals page. Existing links open the Signals overview.
False Positives and Activity Log
Tell Moda when a detection is wrong, and see who changed what in your workspace.
AddedMark any detection as a false positive from the dashboard or the CLI, with an optional reason, and undo it at any time. Marked items leave the product right away.
ImprovedReviews on problem evidence feed back into how Moda matches traces to problems, and rejected evidence stays out of that problem.
AddedOwners and admins can see a workspace activity log in Settings covering settings changes, member role changes and removals, ownership transfers, and API key creation and revocation.
Addedmoda skills install adds bundled integration guides to your coding agent, and moda auth api-key prints the saved API key.
Linked Messages and Emotion Families
Every message behind a finding is highlighted, and steering an agent is no longer counted as being upset with it.
AddedOpening a trace from a problem or custom signal highlights every linked message, with a banner that counts them and arrows to step between them.
ImprovedRedirection, such as "not that one", is shown as a mixed emotion separate from negative emotions, so the negative emotion rate counts only users who are actually unhappy.
Addedmoda signals, moda signal, moda signal-label, and moda detection-review manage custom signals and review detections from the CLI.
FixedSignal evidence links now open the exact message the quote came from.
FixedTool results that return objects are now printed in the trace panel.
Custom Signals for Every Team
Define your own checks for your agents and have Moda read every new trace against them.
AddedCustom signals are open to every workspace. A saved signal goes live right away and starts reading new traces.
AddedSearch past traces for matches before saving, mark hits as correct or incorrect, and explain why when you reject one.
AddedDelete a custom signal from the dashboard.
AddedOwners and admins can mark problem detections as correct or false positive from traces and evidence cards, with a note.
AddedA Settings switch turns live custom-signal detection on or off for the workspace.
Weekly Updates
A weekly write-up of what the Moda team shipped.
AddedPublished the weekly update archive at /weekly, with a permanent numbered URL for each issue.
FixedDistinct messages that share one OpenTelemetry trace are now kept as separate messages.
Whole-Conversation Problem Finding
Moda reads entire conversations to find problems, and the problem view reads top to bottom.
AddedWhen a conversation goes quiet, Moda reads all of it and builds problems and root causes from first principles, so long-running agents are analyzed without losing context.
ImprovedThe problem description now sits under the title.
AddedThe problem Traces tab shows trace excerpts, the reason each one was included, and a link to that moment in the trace.
ImprovedTimeline and World State now sit above the trace.
Faster Long Traces
Long traces open at the turn you were sent to and fill in as you scroll.
ImprovedTrace pages load the page around a deep-linked turn first, highlight that turn, and load the rest as it comes into view.
RemovedRemoved the turn inspector. World State remains on the Memory view.
ImprovedThe Users page explains the health score inline, and the emotion detector's score is labeled Confidence.
FixedInvited teammates now land in the organization they were invited to, existing accounts can be invited, and invitations can be resent.
Users Filters and Silent Tool Failures
Find struggling users in two clicks, and see tool calls that reported success without doing the job.
AddedThe Users page has working filters for health status, health score range, trace count, and tool failures, plus sorting.
ImprovedTool failures separate tools that errored from tools that returned success without doing the job. Silent failures count only when the evidence is unambiguous.
FixedUser Emotion chart tooltips are no longer transparent.
FixedFixed an error that could stop the 24-hour Traces view from loading.
Modesius in the Dashboard
Ask questions about whatever you are looking at, without leaving the page.
AddedAsk Modesius opens a panel beside any dashboard page. Select a chart point, a message, a tool call, or a filtered page, and ask about that one thing.
AddedAnswers cite the exact traces and dashboard data they used, state their confidence, and show verified findings while a longer investigation continues.
ImprovedInvestigations keep running if you close the browser, and follow-up questions keep the original scope.
Problems and Use Cases, Connected
See which recurring problems affect each use case, and which use cases each problem touches.
AddedUse cases list the open problems that affect them, with affected traces and users, daily recurrence, and links to the exact turns.
AddedProblems show the use cases they touch.
Addedmoda problem-impact and the Data API problem-impact endpoint return the same relationships and counts.
Braintrust Import
Bring your Braintrust history into Moda.
AddedConnect Braintrust in Settings, Integrations to backfill project logs from the US or EU region, choosing projects and a date range. Imports run in the background and can be paused, resumed, or canceled.
ImprovedImported traces keep their parent links, timestamps, sessions, tool payloads, and original fields, and flow through analysis as they arrive.
Problems Ranked by Impact
Problems is one ranked list, led by whatever affects the most users right now.
ImprovedProblems is a single list ranked by users affected, with the largest problem in a spotlight card and the next two beneath it.
ImprovedEach problem carries tags for the signal families behind it.
AddedSlack alerts for new problems start a thread with a briefing on that problem, and replies in the thread are answered about that problem only.
ImprovedSlack only alerts on problems that have traces attributed to them.
Website Redesign
A redesigned moda.dev.
ImprovedRedesigned the landing and marketing pages, and restored the customer logo marquee on the homepage.
Conversations Are Now Traces
Moda analyzes every agent run, with or without a human in the loop, so conversations are called traces everywhere.
ImprovedRenamed conversations to traces across the dashboard, docs, CLI, and MCP server. Old /dashboard/conversations links redirect to /dashboard/traces.
AddedModa posts a Slack alert when it finds a new problem. Reply in the thread to ask follow-up questions.
Readable Traces
Code, structured payloads, and harness messages in traces are easier to read.
ImprovedCode, tool inputs and results, and reasoning in traces are syntax highlighted.
AddedJSON, XML, and YAML inside messages, including blobs embedded in prose, are detected and formatted.
AddedTurns injected by your harness are shown distinctly from what the user and the agent said.
FixedThe Full Trace filter counts no longer shift the column when Tool Calls is expanded.
OpenTelemetry Ingestion Improvements
More of what your agents send arrives intact.
ImprovedTool inputs and outputs carried in GenAI execute_tool input and output messages are now captured.
ImprovedNamespaced conversation IDs are accepted throughout ingestion and analysis.
AddedTool-call events accept an agentName so traces can be labeled by agent.
FixedImplausible client timestamps are clamped to ingest time instead of placing messages far in the past or future.
Hosted MCP Server
Query Moda from any MCP client.
AddedA hosted, remote MCP server exposes every remote-capable moda CLI command as an MCP tool. Authenticate each request with your Moda API key.
AddedWorkspace processing settings let you turn individual analysis signals, starting with frustration, on or off.
Interactive Demo and Self-Serve Signup
Try Moda without an account, or create one yourself.
AddedA public, read-only demo of the dashboard with a guided tour.
AddedSelf-serve account creation, with a workspace set up on your first verified sign-in.
Improvedmoda auth commands understand API-key sign-in, and their help text is clearer for coding agents.
Usage Billing
See your usage and invoices in Settings.
AddedOwners and admins can see daily usage, the current billing period, and invoices in Settings, Billing.
Laziness Signal
See where your agent cut corners.
AddedA Laziness tab on the Signals page shows detections such as repeated identical tool calls, premature handoffs, and incomplete work, with a matching Data API endpoint.
AddedNew CLI commands: moda problem opens a problem's dossier and evidence, moda problem-feedback acts on it, moda emotions lists emotion detections, and moda tail streams new traces and detections as NDJSON.
Addedmoda world-state can show what the agent believed at an exact turn or replay state over time, and moda clusters --search finds use cases by meaning.
FixedY-axis labels on the Users charts are no longer clipped.
FixedThe CLI --include-tool-io flag works again, and not-found results now return a warning instead of an empty answer.
Direct Ingestion Parity
The HTTP ingestion API accepts the same detail as OpenTelemetry.
Improved/v1/ingest now accepts the same fields as OpenTelemetry ingestion, so data sent either way is analyzed the same.
Faster Dashboard
The dashboard loads faster from the first click after sign-in.
ImprovedEvery dashboard request is about 450 ms faster thanks to cached session checks.
ImprovedThe last 24 hours of dashboard data are prefetched right after you sign in, and the overview loads faster.
ImprovedRedesigned Use Cases around the workflows your agent handles, and aligned the Users and Tool Catalog pages with the dashboard table design.
ImprovedWorkspace access is checked on the server for every observability API.
FixedJumping between failures in a trace no longer loses your place.
Dashboard Redesign
New navigation, a command bar, and rebuilt Problems and Signals pages.
ImprovedRebuilt the sidebar around your workflow, and made the top bar and Command-K a command surface with recent destinations.
ImprovedSettings is now a full workspace page, and the dashboard overview is redesigned.
ImprovedSignals is rebuilt around detector trends and the evidence behind them.
AddedEach problem has a dossier covering what happens, why, who is affected, how severe it is, the evidence, and how to tell when it is fixed, with emerging issues and data freshness shown on the list.
AddedPublished an OpenAPI specification for the Data API at /api/openapi.json.
FixedMulti-turn OpenTelemetry conversations keep every user turn.
Status Page
Check Moda's availability and incident history.
AddedA public status page at /status with current status, uptime history, and dated incidents.
Problems
Moda groups signals into recurring problems by root cause, so you know what to fix first.
AddedProblems pages in the dashboard rank recurring issues across tool failures, emotions, and other signals by root cause, with evidence and an overview card on Home.
AddedMark a problem fixed, dismiss it, rename it, or flag a wrong attribution.
Addedmoda problems and the Data API problems endpoints return the same ranking and evidence.
CLI Feedback
Report wrong data or CLI quirks without leaving the terminal.
Addedmoda feedback sends a note to the Moda team, optionally linked to the trace, cluster, or tool you were looking at.
Use Cases Redesign
Explore the work your agent handles as a browsable map of tasks.
ImprovedUse Cases is a browsable map with a branch overview, hierarchy browser, task inspector, sample traces, deep links, and mobile layouts.
AddedSearch use cases by name or meaning, and drill into the segments behind any node.
ImprovedUse Cases loads about twice as fast.
ImprovedWorld State facts show the turn where each fact was learned or updated.
User Emotion
Moda detects more than frustration.
AddedThe User Emotion tab replaces Frustration and detects frustration, sadness, confusion, anxiety, trust and distrust, and positive emotions, each with a score and a verified quote, plus a Data API emotions endpoint.
RemovedRetired frustration clustering and the By Root Cause view.
Signals
Frustration and tool failures live together on one Signals page.
ImprovedCombined Frustration and Tool Failures into a single Signals page. Old links redirect.
AddedA back-to-top button on long trace pages.
FixedWorld State facts now appear even when a trace has no user ID.
Reports Retired
The weekly Reports dashboard and snapshot pipeline are removed. Ask and Slack digests remain the intelligence surfaces.
RemovedRetired /dashboard/reports, the Sunday snapshot cron, and /api/reports endpoints. Old report URLs redirect to Ask.
Instant Trace List and Brand Kit
New traces show up right away, and Moda's brand assets have a home.
ImprovedNew traces appear in the trace list as soon as they are ingested, instead of after background processing.
AddedA brand kit at /brand with logos and colors. Right-click the Moda logo to reach it.
Ask Moda from the Terminal
moda ask works without setup and shows its progress as it goes.
Addedmoda ask works with no extra configuration and streams live progress while it investigates.
ImprovedAnswers render as styled terminal output with deep links to the evidence, and long investigations stay connected.
ImprovedRevoked API keys are hidden in Settings by default, and the CLI API key limit is a per-workspace setting.
Fixedmoda auth logout clears the stored API key and workspace binding.
FixedVery large ingested messages are split into smaller queue messages instead of being cut short.
CLI Connect Flow
Connecting the Moda CLI is now a one-click browser approval -- no more typing your password into a terminal-launched form.
AddedOne-click CLI connect: if you are already signed in to the dashboard, approving the CLI is a single Connect button.
AddedGoogle sign-in is now supported when authenticating the CLI from the browser.
AddedCLI logins now create revocable sessions with a 30-day sliding expiry that renews automatically while the CLI is in use.
ImprovedThe terminal login flow no longer involves password entry; the browser handles sign-in and hands the CLI a one-time code.
Conversations Table and World State
A faster, configurable conversation list and a World State view built on real data.
ImprovedThe Conversations page (now Traces) is a single table with a Processed column and columns you can choose.
ImprovedConversation details load progressively, list previews show the actual first message, and the list has an improved loading state.
ImprovedRedesigned World State with live data and resizable panes.
ImprovedThe Tool Failures dashboard and its charts render faster.
Raw Span Audit
Inspect exactly what your instrumentation sent.
Addedmoda audit shows the raw spans behind a conversation or trace, including hierarchy, orphans, and duplicates.
FixedOpenAI tool outputs are captured as tool results.
Support, Privacy, and Terms
Reach the Moda team directly, and read how we handle your data.
AddedA public support form at /support.
AddedPublished the full privacy policy, including the data deletion procedure, and the full Terms of Service.
ImprovedOpenTelemetry spans that do not match a known format are kept as tool calls instead of being dropped.
Search and World State in the CLI
Search your agents' conversations and read their world state from the terminal.
Addedmoda search finds messages by keyword, meaning, or both.
Addedmoda world-state returns a conversation's world state: what the agent knew, what it was working on, and how that changed.
Fixedmoda init no longer loses its authentication when moda.dev redirects through www.
Sign-In Redesign
One page to sign in or sign up, with Google.
ImprovedSign-in and sign-up share one redesigned page, with Google sign-in.
AddedThe CLI checks for a new version once a day and shows an upgrade notice.
Conversation Navigation
Conversation detail pages now support moving directly between adjacent conversations.
FixedFixed an issue where previous and next controls on conversation detail pages stayed disabled instead of navigating between adjacent conversations.
Tool Call Failures Dashboard
The Tool Call Failures dashboard now loads reliably when reviewing recent failed tool activity.
FixedFixed an issue where the Tool Call Failures dashboard could fail to load because the overview endpoint returned 500.
Multi-Day Conversation Context
Long-running conversation traces are easier to follow when they span more than one day.
ImprovedChapter headers now include dates for multi-day conversations, with day dividers inside expanded chapters when the trace crosses midnight.
Conversation Full Trace View
Conversation detail pages now let you switch between the summary-first synopsis and the full chronological trace.
AddedRestored the Full Trace view in conversation details with a Synopsis / Full Trace toggle in the header.
FixedTool failure and frustration deep links now open directly in Full Trace mode so the relevant turn or failed tool call is easier to inspect.
Dashboard Sidebar Contrast
The dashboard navigation reads cleanly now -- section labels and inactive items no longer fade into the canvas.
FixedFixed low-contrast dashboard sidebar section labels and inactive navigation items so they meet accessibility contrast targets in both light and dark modes.
Landing Page Polish
Clearer navigation, smoother mobile behavior, and a cleaner customer proof section on the Moda homepage.
ImprovedIncreased homepage navigation contrast and made the mobile menu animate in-flow so it no longer covers the hero headline.
FixedRemoved the hero image opacity fade that could briefly show a beige placeholder before the horizon image appeared.
ImprovedTightened landing-page section spacing and rebuilt the customer logo grid with consistent visual sizing.
AddedAdded Octolane case-study access from the homepage customer proof section.
Users
A new Users view to see who is using your product, who is active, and who is struggling.
AddedUsers list with KPIs, active users and health-score charts, cohort filters, search, and sort
AddedPer-user detail page with stats, inferred traits, and callouts highlighting frustration and tool failures
AddedConversation activity feed and tool failures panel on the user detail page
AddedIntent distribution card comparing a user's top intents to the tenant average
AddedFrustration trajectory chart with severity-colored cause annotations and 14d delta
FixedResolved an issue where user analytics and clustering refreshes were not updating.
FixedResolved an issue where agent-style conversation views showed only one message instead of the full thread.
Cloudflare Container Pipeline
The analytics pipeline now runs end-to-end on Cloudflare Containers with OpenRouter-hosted models, so embedding, segmentation, clustering, and frustration analysis scale to zero between bursts.
ImprovedMigrated the analytics pipeline from Modal to a Cloudflare Container, with scale-to-zero between jobs.
ImprovedReplaced self-hosted GPU services with OpenRouter API calls for embeddings, summaries, and frustration analysis.
FixedFrustration analysis now runs in a background thread so per-conversation processing returns immediately.
CLI Authentication
The Moda CLI can now authenticate without copy-pasting tokens.
AddedNew /auth/cli page and supporting API for browser-based CLI sign-in and signup.
AddedCLI proxy routes for /api/cli/context and /api/cli/bootstrap so the CLI can hydrate tenant context on first run.
Data API and CLI Docs
The read-only Data API is now reachable from the frontend domain with a documented CLI.
AddedAdded a Vercel rewrite so dashboard.moda.dev/api/data proxies to the backend Data API.
AddedPublished CLI install and usage docs alongside the Data API reference.
FixedConversation list now reports a consistent message count sourced from conversation_logging.
Production MCP Server
Connect Claude, Cursor, and other MCP clients directly to your Moda tenant.
AddedProduction-ready MCP server with shared modules, tests, CI, and Mintlify docs.
AddedMock MCP server for local development and contract testing.
Frustration Detection
Real-time frustration detection runs on every new conversation so root cause attribution is available from the frustration page.
AddedFrustration analysis with root cause attribution runs on each new conversation.
Landing Page Redesign
Refreshed marketing site with environment cards, security section, agent replay, and readability fixes.
ImprovedRedesigned the landing page with environment cards, security messaging, and an agent replay section.
Reports and Cluster Labels
Introduced weekly intelligence reports (later retired) and cluster labels that reflect the full segment population.
AddedWeekly intelligence reports with subagent orchestration and SSE streaming (retired 2026-07-11).
ImprovedMap-reduce cluster labeling now uses every segment summary instead of a sampled subset.
ImprovedServer-side pagination, search, and filters on the frustration page.
Agent Failure Detection
Automatically detect and flag agent tool failures across your conversations.