- Added a multi-tenant SaaS control plane to Mission Control, including tenant records, tenant-to-installation bindings, tenant API keys, and tenant-level admin summaries.
- Added tenant-scoped REST API endpoints under
/api/v1/tenants/...so external CI/CD or webhook systems can read dashboards, list agents/tasks, and create tasks with API-key authentication. - Added a commercial Admin workspace extension for tenant creation, installation binding, API-key issuance, and SaaS readiness guidance.
- Added Docker delivery artifacts (
Dockerfile,docker-compose.yml,platform/bin/deploy/docker_bootstrap.sh) for one-command remote deployment of Mission Control plus OpenClaw runtime bootstrap. - Added
platform/bin/deploy/export_dashboard_postgres.pyto export the SQLite product kernel into PostgreSQL-friendly schema and CSV bundles for larger-scale persistence.
- Expanded Mission Control from a powerful local operator console into a tenant-aware product foundation that can be handed to other teams.
- Reworked tenant summaries to use lightweight task/agent aggregation so multi-tenant admin views stay stable without recursively rebuilding nested dashboards.
- Updated product/runtime version metadata to
1.18.0.
- Added broader frontend internationalization coverage for the high-frequency Mission Control workspaces, including Tasks, Conversations, Themes, Skills, and OpenClaw views.
- Added a mobile-friendly task workspace mode with responsive card layouts so the local product remains usable on phone-sized screens.
- Added a PWA shell for the separated frontend, including
manifest.webmanifest, service-worker registration, installable metadata, and offline static-asset caching. - Added API integration tests plus Playwright end-to-end coverage for login, overview analytics, mobile task rendering, and PWA surface checks.
- Tightened the React frontend delivery baseline by syncing the new PWA assets and test harness into the runtime/tooling flow.
- Rebalanced Vite manual chunking so the commercial frontend no longer emits the previous
>500 kBproduction bundle warning while preserving route-level lazy loading. - Updated setup/runtime version metadata to
1.17.0.
- Added a separated
apps/frontend/application built with React, Vite, and Ant Design so Mission Control now has a real product frontend instead of only inline HTML generated from Python. - Added JSON auth endpoints for the new frontend, including
/api/auth/session,/api/auth/login, and/api/auth/logout. - Added frontend-aware serving mode to
collaboration_dashboard.py, with automaticapps/frontend/distdetection, SPA route support, and/legacyfallback for the previous monolithic dashboard. - Added
platform/bin/runtime/build_frontend.shandplatform/bin/runtime/sync_runtime_assets.shso existing installs can build the new SPA and sync runtime assets without rerunning the full interactive installer. - Added chart-driven overview analytics with a task funnel, agent load distribution, and 24-hour activity trend rendered directly inside the React Mission Control shell.
- Added a relay network visualization to the Activity workspace plus a card/table toggle in Agent Ops so operators can scan coordination and load in more than one view.
- Promoted the local product from a server-rendered single-file console into an API-first backend plus independent frontend architecture.
- Updated Mission Control runtime serving to support local frontend development via CORS-enabled API access, while preserving same-origin production serving after
frontendbuild output exists. - Split the new React frontend by route and moved drawers/modals behind lazy-loaded secondary chunks, reducing the largest shared frontend bundle from roughly
917 kBto about581 kBin production build output. - Deferred global search filtering in the React shell so typing across agent, task, and conversation datasets stays responsive under larger local runtimes.
- Reframed the login screen, product shell, and overview workspace into a more commercial control-plane layout with stronger brand hierarchy, operator summary surfaces, and executive-facing first-screen presentation.
- Added theme-aware frontend language switching and localized the new overview, activity, and agent visualization surfaces for Chinese and English themes.
- Added offline snapshot fallback for
/api/dashboardso the product can keep rendering the latest local operational picture even when the live runtime is temporarily unavailable. - Updated setup/runtime version metadata to
1.16.0.
- Added a managed installation registry to the Mission Control product kernel so one local control plane can track multiple OpenClaw installs.
- Added an installation fleet workspace inside
/admin, including instance cards, local-path registration, and stale-instance removal actions for Owners. - Added installation-registry coverage to
tests/test_dashboard_store.py, extending the automated baseline from users/audit into multi-instance product state.
- Expanded the commercial admin workspace from “seat governance” into the first multi-install control plane, with per-instance theme, router, task-count, and status visibility.
- Promoted the current install to auto-register itself into the fleet registry so the product can always reason about “current instance vs. other managed instances”.
- Added
backend/dashboard_store.pyas the first Mission Control product-kernel module, backed by SQLite and responsible for product-user plus audit-event storage. - Added automatic migration from legacy
product_users.jsonandaudit-log.jsonlintodashboard/dashboard.dbso existing installs can move forward without losing local product state. - Added
tests/test_dashboard_store.pyand a GitHub Actions CI workflow to establish the first automated regression baseline for the product kernel.
- Switched Mission Control account and audit persistence from ad-hoc JSON / JSONL writes to the new SQLite-backed storage layer while keeping the public product behavior unchanged.
- Extended install, validation, and theme-switch runtime script deployment so
dashboard_store.pyis present everywhere the Mission Control app runs.
- Added direct per-agent dialogue entrypoints across Mission Control, including one-click conversation launch from Agent cards, Agent drawers, and a new per-agent launcher inside
/conversations. - Added explicit main-session targeting in the Conversation Center so operators can start from an Agent first instead of hunting through the session list.
- Reframed the conversation workspace from a session browser into an agent-first communication surface, making every configured agent visibly reachable as a dialogue partner.
- Updated the conversation focus summary so the product clearly shows whether you are looking at a real selected session or preparing to talk to an agent's main session.
- Added a Conversation Center inside Mission Control with
/conversations, real OpenClaw session browsing, transcript inspection, and direct in-product dialogue with agents. - Added
/api/conversationsand/api/conversations/transcriptso the product can read native OpenClaw sessions and per-session jsonl transcripts without inventing a parallel chat store. - Added authenticated in-product conversation actions backed by
openclaw agent --json, including transcript refresh after each message.
- Expanded Mission Control from “see + operate” into “see + operate + converse”, so operators can stay inside the product for both task control and live agent dialogue.
- Introduced a dedicated conversation permission for
Owner / Operator, keepingViewerread-only while still exposing real transcript visibility.
- Fixed mixed-output parsing for
openclaw agent --jsonandopenclaw sessions --json, so plugin logs no longer break conversation data or chat actions. - Fixed product refresh behavior so sending a conversation message immediately refreshes the session list and transcript instead of waiting for passive polling.
- Added an OpenClaw Control Center inside Mission Control with
/openclawand/api/openclawsurfaces for native version, schema, gateway, and skills visibility. - Added in-product publishing from local repo skills into the current OpenClaw managed skills directory, so owner users can promote a skill without leaving the product.
- Added managed-skills directory visibility and OpenClaw baseline compatibility signals for
2026.3.12+.
- Tightened Skills Center to focus on repo-local skills while letting the OpenClaw view own native managed-skill runtime visibility.
- Upgraded OpenClaw JSON parsing to tolerate mixed CLI output where plugin or warning logs are interleaved with structured JSON.
- Fixed stale post-action UI state by clearing cached OpenClaw and skill payloads after product mutations such as skill publish, theme switch, or packaging.
- Fixed deep-link support for the OpenClaw workspace so
/openclawnow renders like the rest of the product routes. - Fixed local publish-state detection to work even when
openclaw skills list --jsontruncates discovery on very large managed skill roots.
- Added
platform/bin/install/skill_utils.pyto scan, validate, scaffold, and package Claude-style skill folders using Anthropic's published skill structure patterns. - Added a new Skills Center product workspace in Mission Control with skill catalog visibility, quality signals, scaffold actions, and zip packaging.
- Added a sample
mission-control-release-opsskill to demonstrate progressive disclosure withreferences/and release-oriented workflow guidance.
- Expanded Mission Control from an operations/admin surface into a local skills product as well, with
/skillsand/api/skillsroutes. - Wired the dashboard runtime to read skill metadata from the source repository and surface distribution-ready commands directly in the UI.
- Ensured generated skill scaffolds now pass their first validation instead of emitting malformed YAML frontmatter.
- Ignored packaged
dist/artifacts so skill zip creation does not dirty the repository with generated files.
- Added multi-user Mission Control access with local product accounts, role-based permissions, and Owner Token fallback bootstrap.
- Added a commercial admin workspace inside Mission Control for seat provisioning, role/state governance, password resets, and audit visibility.
- Added secure password hashing plus signed session cookies for local product login.
- Split dashboard admin data into public summaries and authenticated sensitive detail, so static snapshots no longer expose seat rosters or audit trails.
- Elevated Mission Control from a protected local tool to a more commercial-grade product surface with account lifecycle and governance workflows.
- Restored the missing
now_iso()helper so account, audit, and session features no longer fail at runtime. - Prevented the last active Owner from being downgraded or suspended, avoiding accidental product lockout.
- Normalized account status handling so suspended seats are enforced consistently across login, admin UI, and stored records.
- Added an in-product task action studio so users can create work directly from the Tasks module without leaving Mission Control.
- Added task-operation forms inside the task replay drawer for progress updates, blocking, and completion handoff.
- Added authenticated product action endpoints for task creation, progress, blocking, completion, and live theme switching.
- Linked generated installs back to the source repository so the product can invoke
switch_theme.pysafely from inside the UI. - Promoted Mission Control from a read-only dashboard into an operational product surface with toast feedback and action-aware runtime state.
- Returned a non-zero exit code when task creation is rejected, so product actions no longer report false success.
- Allowed navigation to continue while a drawer is open instead of letting the scrim block the left menu.
- Kept generated config metadata in sync with the current release version.
- Added a local login flow at
/loginwith cookie-based session access for Mission Control routes, APIs, and live events. - Added layout controls for collapsing the left menu and switching between operations, focus, and compact product layouts with persisted preferences.
- Turned the Mission Control shell into a protected local product surface with navigation, authenticated app routes, and signed local session state.
- Expanded the top navigation to include menu controls, layout switching, and authenticated sign-out actions.
- Ensured unauthenticated requests to product APIs and SSE streams now return proper auth gating instead of exposing data directly.
- Preserved layout choices across reloads so the product opens in the same working mode the user last selected.
- Added a product-style Mission Control app shell with dedicated Overview, Agents, Tasks, Activity, and Themes modules.
- Added local API endpoints for agents, tasks, events, themes, and deliverables so the dashboard can act as a real product surface instead of a single visual page.
- Added integrated product runbook cards that expose common local commands directly inside the app.
- Reframed
collaboration_dashboard.py --serveas a local multi-view application with navigation, search, deep-linkable routes, and dedicated work areas for operations and delivery. - Expanded dashboard payloads with theme catalog, router context, command palette data, and deliverables inventory to support richer product workflows.
- Kept browser console output clean across the new multi-route app experience.
- Preserved drawer-based drill-downs while expanding the app beyond a single-page overview.
- Added clickable agent inspector drawers so users can open any agent card and inspect live focus, recent signals, and in-hand tasks.
- Added task replay drawers that expose route, TODO progress, and chronological handoff/progress history from both task cards and timeline events.
- Refactored
backend/collaboration_dashboard.pyinto a richer mission-control UI with drill-down interactions while keeping HTML + JSON snapshot generation and the live SSE server. - Expanded dashboard JSON payloads with
taskIndex, replay entries, recent agent signals, and active task cards to support richer downstream visualizations.
- Counted only non-terminal tasks toward agent active-state metrics so completed work no longer leaves agents falsely marked active.
- Suppressed the browser
favicon.ico404 noise in the live dashboard UI.
- Added a live local web panel for the collaboration dashboard with
/api/dashboardand Server-Sent Events (/events) endpoints. - Added real-time client-side updates for the mission-control UI without full-page reloads.
- Upgraded
collaboration_dashboard.py --servefrom static file hosting to a real-time local dashboard server. - Updated collaboration dashboard signatures so change events only fire on substantive task/agent updates.
- Avoided noisy continuous dashboard events caused by timestamp-only changes.
- Preserved static snapshot generation while enabling a live browser panel for the same dashboard.
- Added
backend/collaboration_dashboard.pyto generate an HTML + JSON mission-control view for live multi-agent collaboration. - Added automatic collaboration dashboard generation during install, task refresh, and theme switching.
- Updated the README to surface the collaboration dashboard and visual coordination workflow.
- Extended installation validation to require the collaboration dashboard runtime script.
- Ensured theme switching preserves the visual dashboard capability after agent/workspace migrations.
- Added
platform/bin/install/switch_theme.pyandplatform/bin/install/switch_theme.shso existing installations can switch themes without reinstalling. - Added theme schema validation and shared migration helpers in
platform/bin/install/lib/theme_utils.py.
- Preserved existing channel config, models, memory search settings, and task prefixes when regenerating
openclaw.json. - Preserved existing secrets and gateway token when rerunning
setup.sh. - Tightened default elevated allowlists so new installs no longer inherit wildcard access by default.
- Replaced the old
eval-based theme loading insetup.shwith validated Python parsing. - Deployed
health_dashboard.pyto every workspace and updated validation to check it. - Made
health_dashboard.pysupport--dir,OPENCLAW_DIR, and per-workspace auto-detection. - Replaced the Unix-only
fcntlJSON lock with a cross-platform lock implementation. - Migrated task boards, agent session directories, and workspace artifacts across themes when switching.
- Upgraded the installer and validator to fully provision and verify all 11 agents, including the briefing role.
- Aligned generated
openclaw.jsondefaults with the current OpenClaw 2026.3.12 session and gateway settings. - Added
sessions_yieldto theme tool allowlists and generated SOUL guidance for planner, dispatcher, departments, and briefing roles.
- Fixed
setup.shso the theme task prefix is honored by default and the briefing agent is no longer skipped during installation. - Fixed
validate.shso--dirworks as documented and incomplete installations fail validation. - Fixed
kanban_update.pyso missing tasks return non-zero exit codes instead of logging false success. - Fixed task state transitions to preserve the real responsible department during
DoingandDonephases. - Fixed
health_dashboard.pyso active tasks are loaded from the current router workspace instead of hardcodingworkspace-assistant.
- Verified against OpenClaw
2026.3.12. - Confirmed clean isolated installs for the
corporatetheme.
- Initial release of the Mission Control multi-agent orchestration template for OpenClaw.