The AI Control Plane for Modern Engineering Teams
WOR centralizes your projects, agents, personas, models, tools, MCP servers, tasks, chat, infrastructure, and usage into one workspace. Hermes executes today. Future harnesses can plug in tomorrow.
Bring any AI harness. Keep your workspace, agents, memory, governance, history, and task flow.
AI engineering is fragmented
Your operating system for building with AI is scattered across a dozen tools that do not know about each other. Switch harnesses and you lose all of it.
Agents live in silos
Personas, system prompts, and permissions are trapped inside whichever harness created them. Nothing is portable.
Model routing is ad hoc
API keys, model aliases, and provider fallbacks are copy-pasted into config files no one can audit.
Tools and MCP servers drift
Each machine has a slightly different toolset. There is no registry, so there is no source of truth.
Work intake is invisible
Bugs and ideas arrive in chat, email, and DMs. Nothing routes them to an agent or tracks who is on it.
Spend is a mystery
Token cost lands on five separate invoices. By the time you see it, the budget is already gone.
State dies on switch
Move from one harness to the next and your memory, history, and governance start over from zero.
One registry governs the whole stack
WOR is the control plane that sits above every harness. It holds the durable record of who can do what, with which model, against which repository, on whose behalf — and hands that scope to whatever executes the work.
Projects
Every initiative, its repos, and its members in one place.
Hermes asks. WOR answers. Work ships.
Hermes is the first harness wired into WOR. When it starts a job it requests a runtime profile, and WOR returns the exact scope for that moment — agents, models, tools, tasks, permissions, repositories, and chat routes.
- 01
Hermes requests a runtime profile
A single call asks WOR: for this agent, on this project, right now — what am I allowed to do?
- 02
WOR resolves scope
WOR composes the answer from the registry: model aliases, granted tools, MCP servers, repositories, and chat routes.
- 03
Hermes executes
The harness runs with the right scope every time — no hand-edited config, no guessing, no drift.
- 04
WOR records the trace
Usage, decisions, and outcomes flow back into the registry as an auditable history.
Change harnesses without losing your AI operating system
Hermes executes today. When the next harness arrives, you point it at the same registry and keep everything that matters. Your workspace is the asset, not the tool that happens to run it this quarter.
Your AI operating system
Bring any AI harness. Keep your workspace, agents, memory, governance, history, and task flow.
Every permission, as a graph you can read
WOR models agents, the personas they assume, the projects they serve, the models they may call, and the tools they may use as a connected graph. Grant once, audit anywhere, revoke instantly.
- Scope agents to specific projects, repositories, and models.
- Compose personas from reusable permission sets.
- See the full blast radius of any grant before you approve it.
- Revoke a model, tool, or repository and every agent updates at once.
From a reported bug to a shipped fix
Capture work from anywhere, let a human approve it, assign an agent, and watch it move across the board while chat keeps everyone informed.
A bug or suggestion enters from chat, a form, or the API.
A human reviews scope and impact, then approves the work.
An assigned agent executes with its governed runtime profile.
Chat updates post, the customer is notified, history is recorded.
New bug filed · “Auth token refresh fails”
Awaiting approval · scope: api repo · 1 model
Shipped ✓ · customer notified
Humans stay in the loop, in the channel they already use
WOR routes task events, approvals, and agent progress into Rocket.Chat. Approvals happen where the conversation already lives — no new inbox, no context switch.
- Per-project channel routes for intake, approvals, and ship notices.
- Approve or reject work without leaving chat.
- Notify Ricardo, Tony, and the customer on the same event.
Launch a fully-scoped workspace in one move
Pick a project and an agent, and the launcher hydrates a ready-to-work environment — the right repositories, models, tools, and chat routes already wired in.
- One launch, fully governed: no manual key paste, no config drift.
- Reproducible across every machine on the team.
- New teammates start inside the guardrails on day one.
A versioned registry for the things agents call
Register a model once and reference it everywhere by alias. Register a tool or MCP server once and grant it by name. Change the provider behind an alias and nothing downstream breaks.
Models
MODELS- Provider-agnostic aliases
- Routing & fallback chains
- Per-alias cost ceilings
Tools
TOOLS- Versioned tool definitions
- Grant by name, not by file
- Deprecate without breakage
MCP Servers
MCP- One catalog of servers
- Scoped per agent and project
- Health and version tracking
See the spend before the invoice does
WOR meters token, request, and dollar usage per agent, per project, and per model alias. Set ceilings, watch trends, and stop surprises before they reach finance.
Start as a founder. Grow into the platform.
Founder pricing is a launch window, not the long-term position. Lock it in while it lasts.
Founder
2 users · limited to the first 250 accounts
- 2 users
- Core registry: projects, agents, models, tools
- Hermes runtime profiles
- Work intake & Kanban
- Founder pricing locked while active
Professional
5 users
- 5 users
- Unlimited workspaces & agents
- Model routing & fallback
- Work intake & Rocket.Chat routing
- API access
Team
15 users
- 15 users
- Audit logs & advanced RBAC
- Webhooks & plugin SDK
- Priority support
- Everything in Professional
Enterprise
SSO · on-prem · air-gapped
- SSO / on-prem / air-gapped
- White-label
- Custom support & SLAs
- Dedicated onboarding
- Everything in Team
Questions teams ask before switching to WOR
An AI control plane is a single system that governs the agents, models, tools, MCP servers, tasks, and permissions your team uses to build with AI. WOR is that control plane: it holds the durable record of what each agent may do and hands that scope to whatever harness executes the work.
No. WOR is harness-agnostic. Hermes is the first harness integrated today, and any future harness can request the same runtime profile. You keep your workspace, agents, memory, governance, history, and task flow no matter which harness runs the job.
When Hermes starts a job it asks WOR for a runtime profile. WOR resolves the scope for that agent and project — models, tools, MCP servers, repositories, permissions, and chat routes — and returns it. Hermes executes with the right scope every time, and usage flows back to WOR as an auditable history.
It models agents, personas, projects, models, and tools as a connected graph so you can grant access once, audit it anywhere, and revoke it instantly. You can see the full blast radius of any grant before approving it.
Yes. WOR is a model routing platform with provider-agnostic aliases, fallback chains, and per-alias cost ceilings. It meters token, request, and dollar usage per agent, project, and model alias so you see spend in real time across every provider.
WOR keeps a versioned registry of every tool and MCP server. You register a server once and grant it to agents by name and scope, with health and version tracking — so there is one source of truth instead of per-machine drift.
The Founder Plan is launch pricing at $19.95/month for 2 users, limited to the first 250 accounts. It is a window to lock in early pricing, not the long-term position of the product.
A bug or suggestion enters from chat, a form, or the API. A human approves it, an agent is assigned and executes with its governed runtime profile, and Rocket.Chat keeps everyone informed while the work moves to shipped.
Make WOR your AI control plane
Bring any AI harness. Keep your workspace, agents, memory, governance, history, and task flow.
- Call
- (571) 469-6556
- WOR@mustautomate.ai