Brightwire Agent Runtime
Department-scoped execution: standing role agents, skills, connectors, and a durable surface — always under Agent Governance.
Brightwire Agent Runtime is a department-scoped execution environment: a standing team of role agents (lead + specialists) with shared skills and tools, verification so work finishes or escalates, and connectors on your real operations systems — always under Brightwire Agent Governance.
It is not another chatbot seat, not a generic services “AI shop,” and not unsupervised trading or plant control by default.
How we name an instance: {Department} Agent Runtime — e.g. Market Ops Agent Runtime · Energy Management Agent Runtime.
Why this exists
| Symptom | What breaks |
|---|---|
| Laptop / personal stack | Workflow dies when the champion is offline; no shared ownership |
| Seat sprawl | Many tools; little inventory of who runs what, on which data, at what cost |
| No department grounding | Generic desktop tools don’t live on ETRM, PI/historian, settlements, ISO feeds, or CMMS as first-class systems |
| No durable multi-agent collaboration | One chat window is not lead + specialists with handoffs, review, and escalation |
| Weak production surface | No department web application for the dashboards the desk actually runs on |
| Governance afterthought | Access, cost, audit, and human gates bolted on late — or never |
Runtime addresses the department side. Governance addresses company control. You can adopt Governance while staying on existing seats. Runtime is optional; when you add it, it always carries Governance posture.
Design principles
- Department boundary — one Runtime per named department or workstream; scale by adding Runtimes under one Governance posture.
- Standing role agents — lead + specialists with clear jobs, tools, memory scope, handoffs, and escalation.
- Blueprint, then tailor — start from a proven pack; fit the department’s systems and workflows.
- Governance travels with Runtime — access, cost, audit, and policy are included.
- Dedicated department web application — dashboards and review for that desk, not chat-only.
- Your cloud — production Runtime is an isolated environment in your cloud tenant; Brightwire operates the stack inside it.
- Progressive autonomy — Ask → See → Advise → Act. Trading and plant control stay human-led unless separately scoped.
- Managed operate — Brightwire designs and operates the Runtime with your team.
How Runtime relates to Governance
| Layer | Product | Job |
|---|---|---|
| Control | Agent Governance | Access, cost, audit, policy on supported paths |
| Execution | Agent Runtime | Department role agents + skills/tools + connectors + verification |
| Surface | Department web app (per Runtime) | Dashboards, analytics, and review for that desk |
| Offer shape | What you get | When it fits |
|---|---|---|
| Governance alone | Agent Governance | Sprawl, spend, audit pain; custom multi-agent can wait |
| Governance + Runtime | Governance + one or more department Runtimes | A department needs multi-agent work on market, plant, or ops systems |
| Expand | More Runtimes or deeper actions | Trust and proof are in place |
What lives in a Runtime environment
| Component | Purpose |
|---|---|
| Role agents | Standing personas with scoped instructions, tools, and memory |
| Multi-agent collaboration | Handoffs, review, escalation — work finishes or surfaces to a human |
| Skills and tools | Reusable capabilities under policy |
| Connectors | Integration to department systems of record (scoped) |
| Department web application | Durable UI for dashboards, analytics, and review |
| Team messaging | Human + agent collaboration, including Teams where scoped |
| Knowledge retrieval | Answers grounded in approved documents and sources |
| Identity (SSO) | Enterprise SSO and role-based access |
| Secrets & configuration | Workspace-scoped credentials; values not shown in the UI |
| Observability | Health, events, and usage for operate and cost control |
Trust ladder inside Runtime
Default starting point is not Act. Runtime can host work at any rung; Act only inside pre-approved workflows.
Runtime vs desktop agent tools
| Dimension | Desktop / personal tools | Brightwire Agent Runtime |
|---|---|---|
| Unit of design | One user + one session | Department team of standing role agents |
| Collaboration | Manual tab and chat juggling | Designed handoffs, review, escalation |
| Systems grounding | Paste context / ad-hoc tools | Connectors and skills on ops systems (scoped) |
| Production home | Laptop session | Your cloud tenant + durable app surface |
| Governance | Personal or vendor-seat settings | Agent Governance posture |
| Best when | Individual productivity | Department production workflows with company control |
We do not ask power users to abandon Claude or Copilot. Governance is additive. Runtime is what you add when seats cannot carry department multi-agent work on real systems with production continuity and audit.
- Unsupervised plant or OT control
- Unsupervised trading execution
- One Runtime for the entire enterprise
- Replacing your Microsoft 365 admin or vendor governance stack
Related: Why Governance · How a project runs · Next steps