Mission Control · All browser UIs, one OS

Ensemble, unified.

Every AI you run — Hermes on your laptop, Gemini Ultra, Microsoft 365 Copilot, Claude, ChatGPT, Codex and GitHub — on one board. See how each one connects, what it's best at, open it in one tap, or compose a mission and hand it off.

8Agents & services on the team
1Secure pipe in — Tailscale to your laptop
27Hermes cron jobs forming the automation backbone
3Tiered profiles routing work by cost

Mission composer

Write the mission once, with its deliverable and definition of done. Launch queues it on the mission board below and copies a ready-to-paste brief — dispatch is still manual relay today, so nothing is sent on its own. The plan summary under the form is a proposal you review before anything moves.

Sensitive and Restricted work never routes to free tiers.
Assigned seats — tap to toggle, first seat owns the mission *
Execution planProposal — not yet executed

Review this before launching. Launching queues the mission on the board and copies the brief; the agent still gets it by manual relay.

Nothing is sent automatically — you stay in control of every paste.

Mission board

Launched missions land in Queued. Move them as work progresses — this board is saved in this browser only, until server sync arrives in a later phase.

Round Table

One room for the whole council. This is the static room: each member joins through the real channel listed on their seat, and what they actually said gets pasted into a permanent, append-only transcript. Live multi-agent posting arrives in a later phase — nothing here posts by itself.

Sessions

Start a session with a topic or agenda. Sessions and transcripts are saved in this browser and included in Export / Import.

    Transcript

    No session yet.

      Append-only: once an entry is added it can't be edited or deleted.

      The team

      Specialist roles

      Phase 1 roles — they run on the connections above, routed by model tier, and they're in the composer picker too. Subjects are proposed drafts; correct them and they get locked in. Each role's evidence gate below must be fully checked before one of its missions can be verified.

      How it connects

      One pattern for everything — including services with no API, like Gemini Ultra and ChatGPT on subscription.

      Fluffy · CoordinatorThis conversation — plans, routes, reports back
      TailscaleEncrypted pipe to laptop-agm7mbgn · 100.122.18.91
      Hermes AgentAPI on :8642 · the bridge on your laptop
      Browser & CLI bridgesChrome automation in your logged-in sessions + local tools
      Gemini · Copilot · ChatGPT · Claude · GitHubYour subscriptions, used as-is — no API keys needed for the browser UIs

      Why it works: Tailscale reaches Hermes, and Hermes reaches everything else — the same pattern your copilot-broker prototype already proved for Microsoft 365 Copilot. Gemini runs the identical bridge against gemini.google.com. The Hermes link itself is verified end-to-end; the Gemini and ChatGPT browser bridges are the next build step, so those two run on manual relay (copy-paste) until then.

      Routing & cost tiers

      Who gets what

      The specialty map — Gemini's five are in its own words, from its first Ensemble brief.

      TaskGoes to
      Google Workspace workGemini — Gmail, Drive, Docs, Calendar, YouTube natively
      Live researchGemini — search-grounded, cited, up to date
      Huge contextGemini — whole codebases, PDFs, thousands of notes in one pass
      Video / audio / screenshotsGemini — native multimodal analysis
      Microsoft 365 workCopilot — via the copilot-broker bridge
      Local files & automationHermes — terminal, cron, your C:\tools projects
      Hard reasoning & writingClaude — flagship tier, with Opus as fallback
      Repos, issues, PRsGitHub — via the ghcp profile & Copilot quota

      Hermes tiered profiles

      Already live on your laptop — this is Ensemble's cost-routing model.

      Flagship  default

      GPT-6.1-Sol first, Claude Opus 5 as fallback. For work that has to be right — planning, legal, security, final synthesis.

      Quota  ghcp

      Claude Sonnet 5 running on your GitHub Copilot quota — strong model, already paid for.

      Budget  lite

      Budget Claude Sonnet 5 for bulk and routine jobs — summaries, sorting, drafts.

      The backbone  27 cron jobs

      Morning Briefing, revenue loops, memory mirrors and more — the automations Ensemble orchestrates rather than replaces.

      The free-to-paid ladder

      Every model call enters through one gateway and climbs only as high as the task demands.

      1. LiteLLM gatewaySelf-hosted — the single choke point for every model call, so keys, budgets and fallbacks live in one place.
      2. Free tiers firstGoogle AI Studio's Gemini free tier, Groq's free plan, OpenRouter :free models. Heads-up: free-tier data may be used to train providers — no sensitive work on this rung.
      3. Cheap paid bulkGemini 2.5 Flash-Lite, DeepSeek off-peak, and Llama 3.1 8B-class models carry the worker volume.
      4. Frontier, sparinglyReserved for planning, legal, security and final synthesis — the work that has to be right.
      ApproachEst. / month
      All-frontier, everything~$400
      Tiered — 10% planner + 90% workers~$56
      Tiered + batch jobs~$32

      $500/mo model spend

      Self-reported manual tracker — there's no live billing feed into this page, so type in what each bucket has cost. Saved in this browser only.

      $0 of $500
      $
      $
      $

      Microsoft 365 subscription — $200/yr Billed separately as a yearly subscription; not counted against the $500/mo model cap.

      Ops queue — all clear

      All three laptop jobs are done. Nothing is waiting on you at the machine.