Command Centre Puts Your Tools and AI Agents in One Conversation
June 6, 2026 · 7 min read
There's a version of this that sounds like science fiction: open a chat window, ask "which Stripe invoices over £10k are still unpaid?", get the answer, then say "draft a follow-up email to each of them" and watch it happen. Then realise you do this every Monday morning, and say "automate this" — and next week it runs itself.
That's Command Centre.
It's a single conversational interface connected to your business tools and your AI agent teams. You can query data, take actions, manage what's running, and build new automations — all without leaving the conversation.
Querying your tools
Every tool you've connected to Envelope — Stripe, Linear, Slack, HubSpot, GitHub, your databases — is available to Command Centre. You ask questions in plain English. It reads from the live data and answers.
Some examples of what that looks like in practice:
- "Show me all Linear issues assigned to me that are overdue"
- "What's our MRR this month versus last month?"
- "Which Slack channels had the most activity this week?"
- "Find GitHub PRs that have been open for more than two weeks"
- "What deals moved to Closed Won in HubSpot this month?"
These aren't queries you write. You don't connect to an API or write a SQL statement. You ask the question, and Command Centre decides which tool to call, calls it, and formats the answer.
The answers are live. Ask the same question tomorrow and you get tomorrow's data.
Taking action
Command Centre isn't read-only. Once you have an answer, you can act on it in the same conversation.
"Move those overdue Linear issues to the backlog" — done.
"Send a summary of this week's Slack activity to the #ops channel" — done.
"Create a GitHub issue for each of those open PRs with a reminder label" — done.
The tools are the same ones your agents use. Command Centre just gives you direct access to them, conversationally, without having to build a team first.
The automation pathway
The most interesting pattern we've seen is what happens when a query becomes a habit.
You ask Command Centre the same question every Monday: which deals moved stages last week, what's the support ticket backlog, how many PRs are waiting for review. After a few weeks you notice you're doing the same thing repeatedly. You say: "Can I automate this?"
Command Centre can help you build a team that runs the query on a schedule, formats the output the way you like it, and sends it wherever you want — Slack, email, a dashboard. The conversation that was your weekly manual check becomes an agent team that runs itself.
This is the pathway from ad-hoc queries to systematic automation. You start by asking questions. You end up with a workflow. The tool meets you at whatever stage you're at.
Managing your AI teams
Command Centre also sits above every AI team you're running. Before each conversation, it loads context on your whole workspace: every installed team, their 30-day run history, any pending approval gates, and recent outputs from each team's agent thread.
"Did anything fail overnight?"
"Is there a pending approval gate blocking anything?"
"What did the content team produce yesterday?"
It answers from live data, not a cached summary. If a team is failing consistently, Command Centre can tell you that and show you the pattern. You can also trigger actions directly — run a team, clear a gate, adjust a configuration — without navigating to it.
Pending gates are the highest-urgency signal. In practice, the first question people ask is almost always "is there anything waiting for my approval?" A pending gate blocks a pipeline, and it's the most time-sensitive thing in a workspace. We made gate visibility a first-class part of the context, not an afterthought.
The execution side
The chip suggestions — pre-generated quick actions based on current workspace state — shift the interface from interrogative to exploratory. If a gate is pending, a chip appears. If a team last ran three days ago, a chip appears. If you haven't queried a connected tool recently, suggestions surface. You don't have to know what to ask next — the workspace tells you what's worth looking at.
The context budget is the real design constraint underneath all of this. Fetch too little and the answers are generic. Fetch too much and you've used the context window before the conversation even starts. We landed on structured summaries for workspace data — aggregate counts, status flags, recent output excerpts — combined with live tool calls for connected data. The workspace summary is pre-loaded. The tool queries happen when you ask.
What's next
Cross-tool queries. Right now Command Centre queries one tool at a time. The natural next step is questions that span tools — "which customers who raised a support ticket this month also have an open invoice?" requires both your helpdesk and your billing data in a single answer. That's on the roadmap.
The automation pathway formalised. You can already say "automate this" in Command Centre and it will build the team. What's missing is the moment where Command Centre recognises the pattern itself — "you've run this three times, want to schedule it?" — and offers the transition unprompted.
Command Centre is live in all Envelope workspaces.
Envelope is an open schema and managed runtime for AI agent teams. openenvelope.org
Frequently Asked Questions
What is Command Centre?
Command Centre is a conversational interface connected to your business tools and AI agent teams. You query live data, take actions across tools, manage running agents, and turn repeated queries into automated workflows — all from a single conversation window.
Which tools does Command Centre connect to?
Any tool you've integrated with Envelope — Stripe, Linear, Slack, HubSpot, GitHub, Zendesk, and others. You ask questions in plain English; Command Centre decides which tool to call and formats the answer.
How is it different from a regular AI chatbot?
A chatbot answers questions from its training data. Command Centre answers from your live business data by actually calling your connected tools in real time. It can also take actions — moving data, updating records, triggering agent runs — not just retrieve information.
How does "turn a query into an automation" work?
When you notice you're running the same query every week — checking overdue invoices, reviewing stalled deals — you can tell Command Centre to automate it. It turns the query into a scheduled agent team that runs on its own and delivers results without you having to ask.