What is Volkom Mesh?

A self-contained explainer. First what Volkom Mesh is without the jargon, then the real architecture.

In one sentence#

Volkom Mesh is a platform to build, run and monitor custom automation agents for small and mid-sized businesses — and to show, in business language, what they did and how much it was worth.

We don't sell software the customer has to learn. We sell repetitive work getting done on its own, and the customer watching it happen in a clear portal.

The problem it solves#

SMBs lose hours every day on manual, repetitive work: entering data, reconciling, generating reports, monitoring systems, moving files between applications. The options on the market don't fit:

• Enterprise tools (UiPath, Automation Anywhere): expensive, complex, designed for corporations with an IT team.
• Generic SaaS: doesn't adapt to each business's particular process.
• AI chatbots: they "talk", but don't execute real work.

Volkom Mesh targets that gap: custom automation, managed (the customer installs and maintains nothing) and with visible results.

How it works (simply)#

Three moments:

1. Build. Volkom develops a custom agent for the customer's highest-value process, starting from a shared base (the "carcass") that already has all the plumbing solved. That's why a new agent takes days, not months.

2. Run. The agent runs on its own — scheduled (every X time) or triggered by a multi-step workflow. It runs on the customer's infrastructure; Volkom never has to enter their network.

3. Show. Every time the agent works, it emits telemetry: what it did and how much it was worth. That shows up in a portal designed for business people.

The two things every agent reports#

This is the product's central idea. An agent doesn't just say "it ran fine" — it reports on two separate channels:

The execution channel proves it works, step by step, without errors. The business channel is what the owner actually wants to see: the KPIs of their operation.

What makes it special: the engine is generic#

The portal that shows the telemetry is data-driven: the agent declares what it measures (the label, the unit: money, count, percentage, time) and the portal displays it with no custom code. Adding a KPI or an agent from a new domain never touches the UI.

That's why the same engine serves, without being rewritten, industries that have nothing in common: e-commerce, logistics, healthcare, manufacturing, finance, MLOps, content moderation. That's what makes it scale.

Who controls what#

• Volkom controls the fleet: develops the agents, versions them, deploys them, schedules them, manages credentials and monitors everything from a central panel.
• The customer sees their results: logs into their portal, sees the activity and their KPIs, can pause/resume an agent and rename it — but never touches infrastructure or worries about security.
• Isolation per customer: each customer (tenant) has its own database.
• Traceability: an audit log records who did what on the platform.

In practice: Volkom is the technical team the customer doesn't have to hire.

The names (important)#

• Volkom Mesh = the solution sold to the customer (what the customer sees and uses). "Mesh" for the idea of a mesh of distributed agents running inside many customers' operations, observed from a single place.
• volkom-core = the internal versioned library that powers everything (the engine, the carcass, the telemetry exporter). It's the accelerator: improve once, propagate to every customer on the next version bump.

Mesh is the promise; core is the machine.

Real architecture (for a technical look)#

Two installables + N agents. Push-only communication (the customer pushes telemetry outward; Volkom never enters the customer's network).

┌─────────────────────────────┐         ┌──────────────────────────────┐
│   CUSTOMER PACKAGE           │         │   CONTROL PLANE (Volkom)     │
│   (runs on customer infra)   │         │   (runs on Volkom infra)     │
│                              │  push   │                              │
│  - Orchestrator (Temporal)   │ ──────► │  - Ingest API                │
│  - Agents (workflows)        │  Bearer │  - ClickHouse (telemetry)    │
│  - Telemetry exporter        │  token  │  - Customer portal + Admin   │
│  - Reads local data          │         │  - Alerting                  │
└─────────────────────────────┘         └──────────────────────────────┘

• Customer package (Installable A): orchestrator + exporter + that customer's agents. One build per customer — agents are not shared.
• Control Plane (Installable B): ingest API + ClickHouse + two frontends (the customer portal and the admin panel) + alerting. A single, central one, observing every customer.
• The Core (the accelerator): the reusable skeleton (orchestrator + BaseAgent contract + exporter) as the versioned library volkom-core. Every customer build imports it and adds only its own agents.

Key design rules: - The control plane only observes — it never deploys remotely nor connects toward the customer. Adding an agent is a manual action by Volkom inside the customer package. - Push-only communication: the customer opens only outbound egress, no inbound ports. Auth by per-agent Bearer token (rotation/revocation). - One API and one database per tenant: the customer portal and the admin panel read exactly the same data. No out-of-sync data between frontends.

Scope note (MVP): the "agents" are deterministic workflows (no LLM at runtime); "agent" is the commercial label. MVP agents are headless (APIs, files, databases). Automation that drives a desktop GUI is a separate tier, outside the MVP.

In summary#

Volkom Mesh turns manual SMB processes into agents that work on their own, managed end to end by Volkom, and reports in business language what they did and how much it was worth — on a generic engine that scales to any industry without being rewritten.

For the details of how an agent reports its telemetry, see the Telemetry Guide (explanatory) and the API Reference (technical).