August 2026: Say What You Mean

First real working day back after two weeks off, and none of it went into the product. Compliance calendar, company paperwork, the things that have to happen before anything else can. The agents were building next to me the whole day.

By the evening there was an MCP server sitting on my private mailbox. One pull request. I've run an alias system for years, every shop and supplier gets a forwarding address named after them, so when one of them starts collecting spam I know immediately who sold it. Works great. It also produced a 280 folder archive of a decade that I had never once sorted.

What I didn't do is write the rules first. I opened the inbox, went through it with the agent one pass at a time, and corrected as it went. That's the part I keep coming back to. Not the speed. The correcting.

Here's what shipped.

A Change That Waits for You

Capability editing got a complete review path this month. A proposed change renders as a preview before anything is written, and you approve or reject it inside Outcomet instead of in whatever produced it. Validation and conflict handling follow the same rules at every entry point, so an edit you make by hand and an edit arriving as a proposal fail for the same reasons and in the same words.

Deleting a capability is recoverable now. Delete Undo restores a retired capability, and failed map links or failed proposal actions report what actually went wrong instead of ending in silence. Reparenting without an explicit position preserves the existing sibling order, which sounds like a footnote until you've watched a map reshuffle itself because a drag didn't say where.

Changes an agent makes to the capability graph work the same way. They arrive as proposals that wait for a person, not as writes that already happened.

An agent that rewrites your capability map without asking isn't helping. It's just faster than you at being wrong.

Empty Means Empty

The smallest change of the month is the one worth reading twice. An empty selection is no longer treated as "remove all". It means nobody has decided yet, which is what an empty selection means everywhere else in software.

Theme relationship changes now behave identically in the Outcomet interface and in approved Nova actions, and Nova can explicitly remove feedback and change relationships when that's what you asked for. Archiving a capability also stopped stranding its old links. Stale theme and change links to an archived capability can be cleaned up, while the capability itself stays closed to new ones.

Nova reports the scoped theme-or-evidence error it actually hit, instead of claiming that no feedback chunks matched. A wrong error message costs more than no error message at all, because you go and investigate the thing it named.

The Ticket Comes Back

July shipped the outbound leg, where a theme becomes a Work Item and exports to GitHub or Jira in one click. August added the return.

A GitHub issue can be represented as a first-class Work Item and connected to Changes and Themes in Delivery. Something that started life in a repo can carry a theme and a roadmap position without being retyped as a second record that immediately starts drifting from the first.

Work Items also stopped depending on where you looked at them. The Delivery graph now matches Theme Detail on Work Item relationships, including shared themes, direct theme links, and clearer relationship stacks. Relationship, navigation, and graph behavior agree across detail and timeline surfaces.

Stories Gets the Whole Desktop

Desktop previews don't arrive from the edge anymore. Story and content-source previews open in a resizable in-page detail panel, while smaller screens keep the drawer, and the two widths are remembered separately so setting up one layout doesn't overwrite the other. Escape closes an inline preview and hands focus back to the list or board item you opened it from.

Around that, the surfaces lined up. The Stories canvas reaches the viewport edges instead of stopping at an outer margin. Stories filters apply to the catalog and the in-page detail at once, so a filtered list and the item open next to it no longer disagree. Sources search and filters moved into the same full-width row the catalog already uses, and Source Kanban cards were rebuilt to match the quieter, text-led design Feedback and Themes use.

A Connection That Stays Connected

Connecting an MCP client used to mean pasting a token by hand and logging in again every hour. Codex now authenticates to the hosted Outcomet MCP through the ordinary browser approval flow, with no token to supply, and other OAuth-capable clients can register themselves the same way.

After that it's meant to be something you stop thinking about. The connection carries past the one-hour token expiry without another login, and it survives server restarts and deployments instead of dropping everyone at once.

The tool catalog kept growing behind that connection, mostly ahead of being switched on. Theme audit, theme verification proposals, Work Item lookups, Story tools and typed planning proposals all exist now and are off by default while the review path around them settles. That's direction, not something to go and try yet.

By the Numbers

4 release weeks
13 shipping days
6 marquee modules touched: Delivery, Stories, Sources, Themes, Capabilities, Agent
1 approval gate: no agent-proposed capability change applies without a person looking at it first
1 new connection path: OAuth-capable MCP clients register themselves through the normal browser approval flow

The Details That Add Up

Canonical MCP tool names: duplicate legacy names were retired in favor of one catalog. If you call the tools directly, update the names.
Contact and company write tools: discovery matches what's actually deployed, so the write tools you see are the ones that exist.
Theme Detail: the Work Item link control opens reliably instead of flickering.
Content spacing: tighter defaults in Feedback Inbox and Themes, with an option in Feedback Style settings to let spacing scale with the window instead.

What's Next

July predicted August would push on what happens after the handoff and close the loop back from shipped work to the theme that asked for it. The GitHub issue arriving as a Work Item is that return leg, and it works. What I didn't plan was how much of the month would go into deciding what an agent is allowed to do on its own. That turned out to be the more interesting question, and the answer Outcomet settled on is that it proposes and you decide.

The MCP work continues toward parity. Whatever an agent can do inside Outcomet, a connected agent should be able to do through MCP, and that mostly means finishing the tools that are already half there.

The bigger change is in feedback, where Problems become their own entity. Right now the step from an incoming piece of feedback to a clustered theme happens in one jump, and the thing actually being clustered is a problem that nobody has named yet. Naming it explicitly takes the guesswork out of the analysis, and it shortens the line into Discovery, because a problem is what Discovery wants to start from anyway. Alongside that, smaller work on Contacts, Stories and Campaigns, the kind that only becomes obvious once you're in a module every day.

If you're new here: open a capability proposal and reject it. The interesting part isn't the change. It's that nothing happened until you said so.

— Andy

Full Changelog

New Capabilities

GitHub issues as first-class Work Items: GitHub issues can now be represented as Work Items in Delivery and connected directly to Changes and Themes, so work that originated in a repository can carry a theme and a roadmap position without being re-entered as a separate record.
Native review and approval for capability proposals: Capability proposals can be reviewed, approved, or rejected inside Outcomet, with a preview shown before any change is applied. Capability edits share consistent validation and conflict handling across entry points, failed map links and failed proposal actions return accurate feedback rather than failing silently, and reparenting without an explicit position preserves the capability's existing sibling order.
Delete Undo for capabilities: Capability delete now supports Undo, restoring a retired capability rather than requiring it to be rebuilt.
Agent changes to the capability graph arrive as proposals: Changes an agent makes to the capability graph are created as proposals that require human approval, rather than being written directly.
Explicit relationship removal through Nova: Approved Nova actions can explicitly remove feedback and change relationships from a Theme, with the same rules that apply in the Outcomet interface.
Cleanup of links to archived capabilities: Stale theme and change links pointing at archived capabilities can now be removed, while archived capabilities remain unavailable as targets for new links.
Resizable in-page previews on desktop: Story and content-source previews open in a resizable in-page detail panel on desktop, while smaller screens keep the drawer. Drawer and in-page detail widths are remembered independently, and desktop inline Stories previews close with Escape, returning focus to the originating list or board trigger.
Content spacing option: Feedback Inbox and Themes use tighter default content spacing, with an option in Feedback Style settings to let spacing scale with the window instead.
Connect an MCP client without a token: Codex authenticates to the hosted Outcomet MCP through the normal browser approval flow, with no Bearer token or Supabase credentials to supply by hand. Other OAuth-capable MCP clients can register themselves through the same flow.
Connections that stay connected: An approved MCP connection continues past the one-hour access token expiry without another browser login, and survives server restarts and deployments rather than dropping every client at once.

Changes & Improvements

Work Item behavior aligned across Delivery surfaces: The Delivery graph now matches Theme Detail for Work Item relationships, including shared Themes, direct Theme links, and clearer relationship stacks. Work Item relationship, navigation, and graph behavior stay consistent across detail and timeline surfaces, and Theme Detail's Work Item link control opens reliably instead of flickering.
An empty selection is no longer "remove all": Theme relationship changes behave consistently across the Outcomet interface and approved Nova actions, and an empty selection is no longer interpreted as an instruction to clear every link.
Accurate scoped error reporting in Nova: Nova reports the correct scoped theme-or-evidence error instead of incorrectly claiming that no feedback chunks matched.
Stories and Sources use fuller desktop layouts: Stories canvas backgrounds reach the viewport edges without the earlier outer margin. Stories filters span both the catalog and the in-page detail, and Sources search and filters use the same full-width row above the Kanban and detail split that the catalog already uses.
Quieter Source Kanban cards: Source Kanban cards now match the text-led card design already used by Feedback and Themes.
Canonical MCP tool names: Duplicate legacy MCP tool names were retired in favor of a single canonical catalog. Direct callers should update to the canonical names.
Contact and company write tools match what is deployed: Tool discovery for contact and company writes now reflects the surface that is actually available.

Built, Not Yet Enabled

These exist in the MCP catalog and are off by default while the review path around them settles. They are listed for direction, not as something to enable today.

Theme audit tooling with paginated feedback, evidence, and relationship traversal
Theme verification proposals, which create a review item rather than changing anything directly
Work Item lookup and search for connected agents
Story search, detail, and source tools
Typed planning proposal tools with a human review inbox