Product Update: June 2026 - Context, Converted

Somewhere in the second week of June I went looking for one contact and found three of them. Same person. One row from a LinkedIn import, one typed in by hand months ago, one auto-created off the back of a feedback item somebody filed. Three records, none aware the other two existed.

Small thing on its own, but small things like that are usually a symptom. This one told me the product had grown faster than its own bookkeeping could keep up with.

So June turned into cleanup mode in a few places at once, not because I planned it that way. Contacts and Companies got a real merge workflow instead of a shrug. Campaigns split out from being bolted onto Contacts into its own module, playbooks living in their own studio instead of getting edited mid-campaign. Delivery started keeping its own paper trail with GitHub and Jira instead of leaving that to whoever remembered. None of it looked related while I was building it. By month's end it was the same discipline in different clothes: converge the fact, don't just add another place to write it down.

Here's what shipped.

Campaigns Gets a Room of Its Own

Campaigns used to live as a feature tucked inside Contacts, with playbooks edited live in the middle of a running campaign. That's gone now. Campaigns is its own module, and Playbooks moved into a dedicated Crafting Studio for building stage sets, cadences, and operator instructions before a campaign ever launches.

The module covers more ground than the name suggests. Run a Problem, Solution, or MVP Interview campaign to work through discovery with real people, or run an outbound campaign on LinkedIn or X to test whether a topic gets traction before you commit to it. Same pipeline, different goal. Stage classifications like Success and Cleaning give runs a real shape, spreadsheet and Kanban views cover both ways teams like to look at a pipeline, and a slide-over Member panel handles next actions, touchpoint logging, and turning a member's response straight into product feedback. Success Metrics cards now carry inline sparklines with drill-down into Signals.

The Feedback Inbox Grows Teeth

The Feedback Inbox stopped being a plain list this month. Configurable KPI cards now sit at the top, Open/Closed Feedback, Open/Closed Themes, the Open/Closed Ratio, and a Severity Distribution, and you choose which ones show and in what layout. Source and people linking moved inline: connect a contact or company right from the inbox row without opening a separate dialog, and items can now link to multiple people and companies at once.

Feedback also got heavier. Screenshots and files attach with drag-and-drop and compress client-side before upload, comments turned into threaded discussions with a markdown editor that even accepts a pasted spreadsheet table, and a privacy-preserving Recent Feedback view tags items with Capability Insights so you can scan what's coming in without exposing raw user text. Comment threads are visible to Nova and the other agents too, so a generated summary respects what the team actually discussed.

Records That Clean Themselves

Contacts and Companies picked up a real immune system this month. Creation flows check name and email against active and archived records before letting a duplicate through, inline warnings offer to restore or merge instead, and a built-in merge workflow folds relationships, accounts, campaign links, and activity into one canonical record when two turn out to be the same person or company anyway.

Search got steadier alongside the cleanup. Special characters like commas, parentheses, and backslashes no longer break a query, and every matching contact now shows up in the linking popups instead of some going missing. Archived Companies, Contacts, and Campaigns are badge-visible and restorable with one click, and KPI counters now reflect active records consistently, regardless of which archive filter is active.

A person who finds three versions of the same contact quietly picks the one that looks right and moves on. An agent won't. It reads whatever record it finds first and acts on it.

That's the part that changed how I thought about this stretch of work. Deduplication used to feel like housekeeping, something you do once and forget. Once something other than a person is reading your data, it stops being optional.

The Paper Trail Closes

Delivery started keeping its own receipts. Changes now sync automatically to linked GitHub and Jira items, pushing status, summary, and description updates outward and mapping states correctly, so a closed-but-unmerged PR shows as Cancelled instead of quietly sitting in Done. PR change cards got AI-generated summaries in place of raw commit text, plus capability suggestions you can accept, reject, or re-run.

The Capability Landscape got a matching timeline: a compact calendar view for stepping through history, an import flow that previews CSV/JSON changes before you apply them, and the ability to restore something you deleted without losing the trail. Today's landscape stays fully editable, historical snapshots open read-only. Underneath it all, Vega and Polaris held onto workspace context across turns instead of resetting each time you switched lanes, and a new Testlab section in Workspace Settings let teams opt into or hide experimental surfaces without filing a support ticket.

Outcomet Opens a Door for Outside Agents

June was also the month a second product got born alongside the main one. A new service, outcomet-mcp, went from an empty scaffold to a production endpoint at mcp.outcomet.app in about three weeks. It's a remote MCP (Model Context Protocol) server, the same open standard Claude Code, Antigravity, and a growing list of other agent tools speak natively. Point one at the endpoint, sign in through the same OAuth flow you'd use anywhere in Outcomet, and it can read and act on your workspace with the same permissions a person would have.

The tool surface stays deliberately narrow for now. A connected agent can search and fetch Feedback and Capabilities, and, where a workspace opts in, read Contacts, Campaigns, and Playbooks too. The one write path is the direct line into the Feedback Inbox above, an authorized agent can call add_feedback under a scoped feedback:create permission, but nothing broader. Every request runs through Supabase-backed auth verification, and every auth, access, and write gets logged to an audit trail, so "safe by default" is what shipped, not just a design note. Setup is documented for Claude Code, Antigravity, and other IDE/agent clients, including the Windows stdio proxy handling that made local testing painful until it wasn't, plus a browser-based demo for trying the flow without wiring up a client.

By the Numbers

4 release weeks
23 shipping days
8 marquee modules touched: Feedback, Capabilities, Stories, Campaigns, Contacts, Delivery, Signals, Agents
1 new module: Campaigns, with its own Playbook Crafting Studio
1 new service: outcomet-mcp, a remote MCP server for connecting outside agents
1 new inbound integration: MCP client feedback ingestion

The Details That Add Up

Signal sparklines: inline trend sparklines in the Signal Catalog and a full 30-day sparkline in Signal Detail.
KPI trend dashlets: configurable trend sparklines across Feedback, Theme, and Signal KPI cards.
Planning polish: planning controls consolidated into one vertical list, clearer timeline markers, closed themes visible on the left, and a new Themes/day velocity metric.
Navigation stability: KPI cards and control bars stay put during resizes and when the Agent panel opens.
Graph canvas reliability: Strategy, Signals, Discovery, and Feedback graphs auto-center and auto-fit more reliably.
Stories intake: a Content Sources tab, a drag-and-drop Kanban triage board, and a persistent Quick Capture modal.

What's Next

June made Contacts, Campaigns, and Delivery tend to themselves more than they used to, and it opened a standing door for outside agents through MCP. July likely leans further into both threads: the GitHub/Jira sync reaching deeper into Delivery, Campaigns picking up more of the orchestration Playbooks started this month, and the MCP tool surface growing past its current safe-read, single-write footprint.

If you're new here: open Contacts, run a search with something messy in it, and see if you can find the merge button before you need it. That's the kind of quiet reliability this month was actually about.

— Andy

Full Changelog

New Capabilities

Smarter agents & persistent workspace context: Vega and Polaris agents now start conversations with a compact strategy & discovery context and retain lane-specific workspace context across turns. That means agents surface relevant opportunities, hypotheses, and experiments earlier and stay context-aware across a session, reducing repeated setup and helping you act faster on suggested next steps.
AI-driven PR summaries + capability suggestions: Pull request change cards and detail views now surface AI-generated summaries (when available) instead of raw PR text, and offer contextual capability suggestions you can accept by linking to existing capabilities or creating suggested new ones. Teams also get explicit controls to re-run suggestions or reject them, so automation accelerates triage without taking control away from humans.
Feedback-to-workflow automation: Changes in Outcomet now synchronize automatically with linked GitHub and Jira work items: status, summary, and description updates flow to connected tickets and transition their workflow states to match change lifecycles. Status mapping has been improved so cancelled/closed-but-unmerged states map sensibly (e.g., Cancelled / Won't Do), reducing incorrect ticket states and manual reconciliation.
Capability Landscape timeline & imports: A compact, responsive Capability Landscape timeline gives teams calendar pick, play/step-forward playback, and "Jump to Now," enabling clear historical review and scenario exploration. Timeline snapshots surface capabilities (including soft-deleted ones) as they existed at the chosen date while keeping current management focused on active items. Today's landscape remains fully editable, historical snapshots open read-only to protect the record. The new unified import dialog (CSV/JSON) supports a minimal JSON format, previews proposed graph/list changes, lets you selectively apply updates, restore deleted capabilities, blocks unresolved parents/cycles up front, and includes an "Undo last import" so large imports are safe and reviewable.
Automatic Theme Learning & richer feedback insights: Theme learning is now automatically generated in the background whenever feedback chunks are linked or unlinked. Synthesis prioritizes chunk-level content and severity to produce more accurate, actionable learning cards and surfaces clear updating states during synthesis. Complementing this, new dynamic KPI cards (Open/Closed Feedback, Open/Closed Themes, Open/Closed Ratio, Severity Distribution) provide at-a-glance trend tracking with user-configurable layouts, and configurable trend dashlets carry the same sparkline treatment across Feedback, Theme, and Signal KPI cards workspace-wide.
Feedback attachments, comments, and richer editing: You can now attach files and screenshots to feedback and learnings via a modern drag-and-drop uploader with client-side image scaling/compression, and compose threaded discussions using a new rich-text comments editor that supports markdown preview, raw mode, keyboard shortcuts, and pasting/editing tables. Comment threads are available to AI assistants so generated summaries and exploration respect team discussions.
Stories: Content Sources + Kanban triage: Stories get a dedicated Content Sources tab and an Inbox / Kept / Dropped Kanban for fast triage. Quick Capture is centralized in a header modal and URL capture now produces richer link previews and post descriptions (better LinkedIn handling). The Story Editor includes an Inbound Content Sources panel so anchor and supporting content can be linked directly to a draft.
Campaigns & Playbooks (new separation + tooling): Playbooks and Campaigns are now distinct. The Playbook Crafting Studio lets you build playbook templates visually (pipeline stages, timing, operator instructions). The new Campaigns module provides customizable pipelines, dual Spreadsheet/Kanban views, improved campaign member sheets, and a Campaign Member panel for next actions, touchpoint logging, and converting responses into product feedback. Campaign Success Metrics cards carry inline sparklines and drill down directly into Signals, so campaign outcomes are measured against product signals in the same view.
Simplified external auth & hosted demo: External users can connect using a hosted OAuth/PKCE flow without needing API keys or workspace IDs, and the Hosted Demo reliably launches OAuth sign-in when production demo mode is enabled. These changes reduce friction for external collaborators and demo scenarios.
Privacy-preserving partner feedback (MCP): For partners using MCP, teams can view recent feedback as anonymized summaries and receive structured, capability-level MCP summaries, surfacing trends and prioritization without exposing raw user text. MCP clients can also request permission to submit feedback directly into the Outcomet inbox via a managed, permission-based flow (the add_feedback tool, scoped to feedback:create).
outcomet-mcp: a remote MCP server for outside agents: A new standalone service at mcp.outcomet.app lets external agent clients read and act on a workspace over the open Model Context Protocol standard. OAuth authorization-code + PKCE sign-in means a connected agent inherits the same identity and permissions as the workspace it connects to. Safe-read tools cover Feedback and Capabilities by default, with optional Contacts, Campaigns, and Playbooks reads behind workspace-level flags. Every auth, access, and action event is recorded to an audit trail. Setup is documented for Claude Code, Antigravity, and other IDE/agent clients, including Windows stdio proxy handling, plus a browser-based demo for testing the flow locally.

Changes & Improvements

Feedback inbox & UX rework: The Feedback Inbox was reorganized for scanability and scale: clearer source filtering, improved active-filter behavior and chips, a more compact filter sidebar grouped chronologically, and refined feedback card metadata and layout to surface reporter, source, and counts more clearly. Multi-source linking now supports associating a feedback item with multiple people and companies, and linking/unlinking updates views immediately.
Signals, sparklines & trend visuals: Signals now include inline sparklines in the Signal Catalog and full-size 30-day sparklines in Signal Detail. Historical trend visuals were made more accurate (respecting metric polarity and preserved 0% states) and signal configuration was standardized across modules with a shared Configure Signals modal and keyboard navigation.
Planning, timelines & velocity: Planning and roadmap controls have been consolidated into a single vertical control bar and a dynamic layout mode was introduced to reduce overlap and improve readability. Timeline markers now align with Now/Next, show relative-duration markers for closed themes, emphasize weekly markers, and display closed themes to the left to expose historical throughput. A new velocity metric (Themes / day) is available as a Dashboard KPI to quantify roadmap delivery pace.
Capability health & evolution: Capability health visuals were expanded with trend dashlets, sparklines, and aggregate cards. Evolution statuses became configurable in Workspace settings and are now consistently applied across graphs and detail views. History mode allows applying day-specific Health/Evolution overrides without changing live fields, and playback behavior was smoothed for reliable scrubbing.
Workspace settings & Testlab opt-ins: A new Testlab section in Workspace Settings centralizes experimental/beta feature toggles (Discovery Learnings, External Connections, Feedback Graph) so teams can opt in selectively. Discovery UI now ties hypothesis creation to Validation Tree context (removing the top-level New Hypothesis button) to keep creation connected to the relevant bet, and Testlab changes warn on unsaved navigations.
Navigation, global controls & keyboard support: Page controls and KPI configuration were standardized across the app with contextual overflow menus and number-key quick tabs. Global keyboard shortcuts and in-graph shortcut help were added to speed navigation and reduce clicks.
UI and rendering improvements: Graph canvases and visual surfaces now resize, center, and auto-fit more predictably across panels (including the Agent panel), reducing clipping and layout jumps. Kanban columns and long lists use lazy loading for better performance on large datasets, and timeline/capability graphs initialize centered so historical and live views feel stable during interaction.
Consent & connection clarity: Consent approval flows now surface clear errors when required account or workspace context is missing, and Agent Connections show an authorization-ready state with clearer guidance so teams understand how to complete integration setup.
Delivery controls: Delivery change details now surface manual controls to re-run summaries/capability suggestions and an explicit reject action for suggested summaries/capabilities, giving teams direct control over automation applied to delivery changes.
Duplicate prevention & record merging (Contacts and Companies): Creation flows check name and email against both active and archived records before letting a duplicate through, with inline warnings that offer to restore or merge instead of a silent block, plus a safe "Create anyway" override. A built-in merge workflow consolidates relationships, accounts, campaign links, and activity into one canonical record. Archived Companies, Contacts, and Campaigns are badge-visible and restorable with one click.
Contacts data integrity: Contacts page KPI counters now reflect active records consistently, regardless of which archive filter is applied. Search handles special characters (commas, parentheses, backslashes) without erroring, and feedback-to-contact linking reliably surfaces every matching contact in linking popups and sidebars.