Org Studio

North Star

Org Studio is where hybrid human + AI teams run their work together.

AI agents are becoming real teammates — not tools, not automations — and real teammates need real org scaffolding: clear ownership, a vision to work toward, guardrails to work within, culture, and feedback loops. The answer to AI-augmented work isn't more tools with AI sprinkled on top — it's fewer tools that treat agents as first-class participants from the ground up.

The atomic unit of Org Studio is the owned domain: a teammate (human or agent) accountable for an area, working toward a vision within guardrails, measured by feedback loops. Vision, work, culture, and measurement all orbit ownership. Conversation is a feature of a well-run org — not its foundation.

Open-source core. Runtime-agnostic. Solo developer to enterprise. BYOA (bring your own agent) always free.

Three Pillars

Ownership as the primitive. Every project area has an accountable owner with a domain contract (ORG.md), a roadmap, and guardrails. The product makes ownership, autonomy, and accountability legible — who owns what, what they're working toward, and how it's going.
Autonomy within guardrails. Today's agents need handholding and spoon-feeding. Org Studio's bet: define the vision, roadmap, culture, and team — then let agents work autonomously inside an explicit leash. The leash is autonomy.approvedThrough (agents execute any task ≤ horizon, never above; humans move the horizon). The ownership tenet pushes agents to continuously analyze gaps and opportunities and autonomously evolve their owned domain — gated, when a version declares a measurable goal, on the metric being met (outcome-bound versions), not just child-ticket status.
Turnkey teams. A library of teammate archetypes — each with default skills and a default runtime (OpenClaw or Hermes) — lets anyone staff a domain in seconds instead of configuring from scratch. (e.g., Sam the lawyer · Kate the GTM specialist · Billy the fullstack eng.) Persona archetypes are also the delivery vehicle for vertical org templates (SMB / startup / enterprise).

Current Status (audited 2026-05-29 against HEAD 1059d40)

Shipped & live

Autonomy horizon (approvedThrough) — agents execute ≤ horizon, hard-stop above; auto-advance within. Wired across scheduler, vision, roadmap, store APIs.
Outcome-bound versions (#1263) — successCriteria / metricCurrent / metricTarget / metricComparator / loopPaused on versions; completion gates on the metric, not just child tickets. Shipped & tested (version-metric.test.ts).
Multi-tenant schema — org_studio_workspaces + org_studio_workspace_memberships tables, workspace-auth.ts middleware, /api/workspaces, isolation tests. Shipped.
Per-agent API tokens (#1383) — api-tokens.ts, /api/admin/tokens, tests.
Semver + approval-horizon convention — decision doc present; but repo does not actually follow one scheme — see Known Issues / #1560.
Telegram demoted to health-alerts bridge — comms relay OFF by default (ENABLE_TELEGRAM_COMMS).
Offline draft queue — useOfflineQueue, drafts persist + flush on reconnect.
Activity feed + Performance dashboard — real-time feed on home; /performance per-agent metrics.
Classic dashboard IA — sidebar route group: projects, tasks, agents, cron, calendar, memory, vision, team, performance, health, activity, settings.

Reverted (do not claim as shipped)

Thread-as-universal-primitive + mobile-first 3-tab IA (the v0.14 pivot) — built, then fully reverted (commit 3c86dac "remove chat/DM feature"). ChatThread, DmSidebar, /dms/*, tabs/ routes, and the proto-mobile prototype are all gone. It crippled the structured affordances (roadmap, board, vision doc) by flattening everything into chat — fighting its own North Star. Demoted permanently from a pillar. May be revisited ONLY as an opt-in comms feature within a domain, siloed behind a new-experience flag, never as the product's foundation.

Vaporware in the old doc (claimed, never built)

Web push notifications — no service worker, no pushManager, zero push code. A PUSH_NOTIFICATIONS flag exists with no consumer. Either schedule for real or cut the claim. (This revision cuts it from "shipped.")

Dead feature flags (toggles with no consumer) — RESOLVED 2026-06-02

MOBILE_FIRST_UX (IA deleted), PUSH_NOTIFICATIONS (no push code), TELEGRAM_MIGRATION (script never authored) all rendered in Settings but did nothing. Removed in the Cloud Prep dead-flag cleanup (commit cdc7c26) — the generic flag infra was kept for future experiments. The Telegram migration claim behind that flag was formally dropped (see Cloud Prep below).

Conventions

Ownership is the primitive. Tasks, roadmaps, vision sections, and metrics all hang off an accountable owner + domain. Any feature that doesn't sharpen ownership/autonomy/accountability is suspect.
One leash. autonomy.approvedThrough per project is the single autonomy control. Agents execute ≤ horizon; humans move it. No parallel auto-approve flows.
One field per concept. If the product grows a second flag/column describing the same thing as an existing one, simplify before shipping.
No dead toggles. Don't ship a feature flag without a working consumer behind it. (Violated today — see Dead feature flags.)
Versioning — two deliberate schemes, never crossed (#1560):
- Repo / skill releases → semver (package.json, git release tags). Answers "is this a breaking change to the published artifact?"
- Product roadmap + autonomy horizon → calver (YYYY.MM, e.g. 2026.05). The store and autonomy.approvedThrough use this. Answers "how far along is the product, and how far is the agent leash extended?"
- These answer different questions; conflating them caused the version mess. Repo semver and roadmap calver are independent and that is correct.

Forward Arc

Milestones are named; calver targets are the roadmap anchor (not v0.x labels — that numbering is retired per #1560). "Cloud Launch" remains the qualitative v1.0-equivalent gate: bet-your-business ready.

Cloud Prep (current, ~2026.05–06). Multi-tenant ✅, per-workspace auth ✅, per-agent tokens ✅. Dead-flag cleanup + version-scheme reconciliation (#1560) closed out here. Telegram migration script formally dropped (2026-06-02): no script was ever authored and there is nothing to migrate — agents use Org Studio + Telegram-as-notifications directly (the old comms relay was already demoted to an optional health-alerts bridge, #734). The claim is retired rather than built. Onboarding flow polish + pre-launch security audit moved OUT of Cloud Prep (2026-06-02, Basil): they're launch-gating, not prep — relocated to the front of Cloud Launch + Billing. Cloud Prep is effectively complete.
Autonomous Domain Evolution (Pillar 2 made real). Promote outcome-bound versions from "shipped primitive" to a product story: manual→assisted metric capture, the experiment-loop UX, the "all tickets done but metric unmet → propose next experiment" nudge surfaced in the UI, loop safety caps visible. This is the moat — the thing competitors don't have.
Teammate Library (Pillar 3 made real). Persona archetypes catalog with default skills + runtime. One-click staff-a-domain. Showcases runtime-agnosticism as a visible feature. Foundation for vertical templates.
Cloud Launch + Billing (the commercial-readiness gate). Opens with the two launch-gating items relocated from Cloud Prep (2026-06-02): onboarding flow polish and the pre-launch security audit (auth surfaces, per-agent tokens, CORS, workspace isolation, public-endpoint exposure) — both are entry gates to this milestone, done before public open. Multi-tenant cloud opens publicly. Free tier (BYOA, 1 workspace). Paid tiers delivered as vertical org templates (SMB: accounting/legal/marketing/support · startup: MVP/design/GTM/CS · enterprise: TBD) — curated persona + roadmap bundles on top of the free core, not just seat counts. Legal + ops (LLC, ToS, privacy, GDPR, SLA).

Commercialization (reframed)

OSS core stays free forever — MIT, BYOA, runtime-agnostic. This is the top of the funnel, not a loss leader to be paywalled later.
Activation = Teammate Library. A new workspace is useful in 60 seconds because you staff it from persona archetypes, not a blank board.
Monetization = vertical org templates, not seats. Paid tiers are curated bundles of personas + roadmap scaffolding per segment (SMB / startup / enterprise), sitting on the free BYOA core.
Retention = autonomous evolution. Owned domains that self-improve within guardrails are sticky in a way that a task board isn't.

Designed but Unscheduled

(Outcome-bound versions moved OUT of this section — it shipped. See Current Status.)

Thread-as-comms-within-a-domain (opt-in, flagged) — only if a clear need surfaces.
Native mobile apps (post-v1.0 parking lot).
Plugin system, calendar integration, AI task decomposition, agent self-awareness API — post-v1.0 parking lot.

Boundaries

Fewer tools, not better tools — every feature must replace a tool or deepen an existing surface.
Don't rebuild Slack. Conversation is a feature inside a domain, never the product's foundation.
No production deployments from Org Studio — it's the org layer, not CI/CD.
No direct DB access — all data through store API.
Runtime-agnostic by design — OpenClaw adapter optional.
Keep the store file-backed for local mode — Postgres optional.
MIT license — no proprietary features in the OSS core.
Versioning — two deliberate schemes, never crossed (#1560): repo/skill releases use semver (package.json, git release tags); product roadmap + autonomy.approvedThrough use calver (YYYY.MM). Independent by design.
Stay pre-launch indefinitely until commercial readiness — "Cloud Launch" = "bet-your-business ready," nothing before.

Change History

| Date | Version | Author | Change | |------|---------|--------|--------| | 2026-06-02 | 2026.05 | Basil/Mikey | Reality audit + North Star reframe. Full codebase audit (HEAD 1059d40) corrected major drift: thread-as-primitive / mobile-first IA were REVERTED (not shipped) — demoted from a pillar; web push was never built — claim cut; multi-tenant schema + outcome-bound versions (#1263) were SHIPPED — promoted out of blocked/unscheduled. Reframed the North Star primitive from thread to owned domain. Established three pillars (Ownership / Autonomy within guardrails / Turnkey teams). Scheduled v0.17 (Autonomous Domain Evolution) + v0.18 (Teammate Library). Reframed commercialization around vertical org templates over seat tiers. Flagged 3 dead feature flags + the four-way version-scheme collision (cleanup ticket #1560). |