Command Center
Recommendations
Effectiveness — did past auto-created tasks help?
Control Room
Objective Dispatcher
Runs one objective through the SDLC pipeline — CTO → requirements → architecture → coding → review → tests → QA → sandbox verification — and opens a pull request that waits for your approval. Nothing is merged automatically. One run per organization at a time.
Agent Desks
Active SDLC Queue
R&D Inbox
Trend radar — latest findings (untrusted third-party content)
Model Training Flywheel
Master Plan Roadmap
Ask NEXORA
Type an objective in plain language. By default NEXORA picks the best-fit active agent for it automatically (matched on role/department against your objective) and runs it through that agent's full tool-using loop — the same path a workflow's AGENT_TASK step uses — and shows you what it did and who it picked. Override the pick below if you want a specific agent instead.
Voice is opt-in. Audio is sent only to NEXORA’s private local faster-whisper + Silero service, held ephemerally during transcription, and never stored. Review or correct the transcript before Send. Playback uses local Kokoro and can be interrupted.
Conversation
MCP Hub
Connect MCP servers once; every agent reaches them through mcp_list_tools and mcp_call_read_tool. Tools you mark read-only run directly; every other tool only runs through an approved workflow step (mcp_call_tool). Credentials are stored in the vault, never here.
Connect a server
Connected servers
Agent Skills
Skills are instructions added to agents' prompts, per role (sdlc-coder for the coding stage, * for every agent). Agents propose improvements from lessons learned; nothing changes until you approve it.
Proposals awaiting your review
Import skills from GitHub
Reads <folder>/<skill>/SKILL.md (the Claude skill layout, e.g. affaan-m/ECC (everything-claude-code), folder skills). Imported skills are on-demand: agents see a list, the text is added when it matches the task, and agents can read one with skill_read. Importing gives all active agents skill_read.
Active skills
Add or edit a skill
Knowledge & Obsidian
Search NEXORA's existing indexed knowledge or import/export Obsidian-compatible Markdown. Paths are metadata only; the server never reads your vault filesystem.
Hybrid knowledge search
Indexed documents
Import Obsidian Markdown
Agents & Workforce
Organization Map
live people → agents → workers → handoffsAgents
Create Agent
Workforce Workers
Full Workforce Roster (admin — every registered worker, including gated-off ones)
Every worker registered in this deployment's WorkforceRegistry, whether or
not it's currently admitted — the "Workforce Workers" table above only shows what's live right now. A
PENDING/BLOCKED row here is a department with no working worker yet, or one waiting on an admission
flag/credential; see src/workforce/bootstrap.ts for what gates each one.
Worker Nodes
Machines (the NEXORA server, your PC) that run your logged-in Claude Code / Codex / Gemini CLI
subscriptions as workers. Add a node, copy its token once, and start
ops/worker-node/nexora-worker-node.mjs on that machine. Admission still needs
ENABLED_WORKFORCE_WORKERS + WN001_CLI_WORKERS_APPROVED.
Recent Assignments
Workflows
All Workflows
Execution History
Software Engineering
Repositories and changes managed by NEXORA's own code-quality/testing/deployment/
security-review/architecture agents (Phase 8, SE-001..005) via /api/engineering.
Repositories and changes are created by agents/CLI against a real git worktree, not from this
panel — this is the browse/gate surface: inspect artifacts, check merge readiness, reconcile
against GitHub.
Repositories
Release autopilot
Verified changes the autopilot carries from pull request to production: CI, merge, deploy, health check, and automatic rollback. Changes that touch deploy, CI, migrations, security or dependency files always wait for you.
Changes
Marketing & Social
Launches a content campaign as a real WorkflowDefinition
(research → SEO → social draft), reusing the Phase 9 MKT-001/002/003 substrate and the
existing Workflow engine — no new orchestration. Track a launched run in
Approvals.
Launch a content campaign
Research is dispatched to the local-ollama-reasoner
workforce worker (capability content-research) — inert until that worker is admitted
in production. SEO and drafting run through the selected agents' own AgentRuntime loop.
Campaigns
Media Studio
Launches the Phase 10 YouTube production pipeline (script → SEO → thumbnail →
approval → publish) as a real WorkflowDefinition, reusing
createYouTubeProductionWorkflow() unchanged. Publish is human-approval-gated —
decide it from Approvals.
Launch a video production
Script/SEO dispatch to local-ollama-reasoner
(capabilities youtube-script/youtube-seo); thumbnail generation and
publish are credential-gated connectors (google-imagen, YouTube) — inert until
provisioned, same as every other side-effect connector in this deployment.
Productions
Monitoring & Incidents
System Status
Incidents
Alerts
Anomalies
Intelligence & KPIs
Optimization Recommendations
Forecast
Approvals
Pending Approvals
Recently Triggered Runs
Look up a run
Usage & Cost
Workforce Usage Today
Cost (current window)
Settings → AI Providers
Credentials are write-only and encrypted in NEXORA Vault. Fallback is off unless explicitly enabled.
Model routing
Explicit fallback policy
Settings → API Access
Issues the bearer key external OpenAI-compatible clients (e.g. VS Code's Custom
Endpoint provider) use to authenticate against /v1/chat/completions. The key is
generated and stored in NEXORA Vault and shown here exactly once, at generation time — it
cannot be displayed again afterward. Generating a new key immediately invalidates any previous
one for your organization.
WF-021 bridge key
Team & Roles
Members
Add Member
No email/invite provider is configured on this deployment, so a new member can't set their own password yet — after adding them, use "Set Password" on their row to give them a working login (they'll be required to change it on first sign-in).