Home xeromike xeromike
Get started

Xeromike-agent

Digital employees you can talk to, configure and share

Xeromike-agent is not a one-off chat bot. It is an AI employee system that runs online and can be customised as far as you need: pick employees the way you pick contacts, open a session, map tools and workflows — or enable an employee that is already prepared in the market and start working immediately.

  • Several manageable digital employees, not a single chat box
  • Model, tools, MCP, skills, SOP and prompts configured online
  • Tenant-isolated data in the cloud — your local files are never touched

Enable an employee from the market Create your own employee

Capabilities

Conversation and configuration in one workbench

Daily use feels like instant messaging. When something has to be tailored, you never leave the conversation to manage employees, sessions and resources.

Conversation workbench

  • Contact-style employee list

    The left panel lists every AI employee available in the current project — your own employees mixed with the shared ones you have enabled — with avatar, name and description.

  • Instant-messaging style chat

    Select an employee and the chat opens on the right: several sessions, message sending, attachments and streaming replies.

  • Named sessions, session-level prompts

    One employee can hold many sessions. A session name is shown when you set one, and each session can carry its own system prompt without interfering with the others.

  • Manage the current employee in place

    Your own employees: edit the profile, bind resources, manage tools, authorise who may access them, update the service, edit the prompt, delete. Shared employees: set your own prompt, or stop using them.

  • Company documents and scheduled tasks

    Reference files from the company document space inside a conversation, and create scheduled tasks so an employee keeps working to a plan.

Create and manage your own employees

  • Create an AI employee online

    Fill in name, introduction, avatar, model and run node — and a conversation-ready employee exists.

  • Multilingual display

    Employee names and introductions support several languages, so users in different regions see the matching copy.

  • Clear permission boundary

    Say which users may access an employee. Employees belong to the current project, so switching project switches the whole set of employees and configuration.

  • Connected to business channels

    Hosted company email, WhatsApp, LINE and AI Chat can be bound to a given employee and session, so outside messages are received by the right digital employee.

Runtime and usage

  • Employees run on online nodes

    Conversation, tool calls and session state all go through the online service, so an employee keeps running on its deployment node.

  • Tokens and usage you can check

    Public nodes are billed by plan and token top-up, and usage is visible inside the project, which keeps cost accounting simple.

Employee market

Start with an employee somebody already configured

The employee market is a separate entrance for discovering and using shared AI employees — ideal for "use it first, then decide whether to build your own".

  • Marketing · Email Management Robot

    Sorts, labels and drafts replies for shared inboxes, and escalates whatever needs a human.

    Use

  • Marketing · WhatsApp Robot

    Answers WhatsApp enquiries in your tone, qualifies the lead and hands the conversation over when a person is needed.

    Use

  • Promotion · Web Posting Robot

    Turns your notes into posts and publishes them to a schedule you keep control of.

    Use

  • Financial Management Robot

    Reconciles statements, flags anomalies and drafts the monthly close for review.

    Use

The four listings above are examples of the kind of employees the market carries; what is available at any time depends on what publishers have shared.

  • Browse reviewed shared employees

    The market shows avatar, name and introduction so you can judge quickly what scenario an employee fits.

  • One click to use, one click to stop

    Once enabled, the employee appears in your conversation list and can be chatted with just like your own; stop using it and it leaves the list again.

  • Write your own prompt

    On a shared employee you can write the system prompt that applies to your own use only — the publisher's original configuration is never changed.

  • Publishers can share too

    Your own employee can be set to shared; after review it enters the market for other tenants to discover and use, settled on the billing terms agreed at publication.

The market answers "where does the capability come from": without building models, tools and workflows from zero, you can own a working AI employee right away.

Shared resources

Employee capability assembled from reusable resources

Xeromike-agent breaks employee capability down into a reusable resource library. A resource can be private or shared, and it takes effect once it is mapped to a specific employee.

  • Tool

    Tools

    Let an employee act: call system capabilities or external interfaces.

  • MCP service

    MCP services

    Connect external systems and services over a standard protocol, widening the set of tools an employee can call.

  • Skill

    Skill documents

    Add reusable professional skills and knowledge packs to an employee.

  • SOP

    Execution workflows

    Freeze standard operating steps so an employee follows a process instead of relying on an improvised brief every time.

Using them is direct

  1. Create or pick tools, MCP services, skills and SOPs in the resource library (my private / shared).

  2. Map the resources you need onto one AI employee.

  3. Go back to the conversation and verify: the employee now works with the mapped capability.

One resource can be reused by several employees, so changing employees never means reinventing tools and workflows. The shared employees in the market are built on this same reusable resource system underneath.

Characteristics

Five things that set Xeromike-agent apart

Nodes, customisation, authorisation, data and the entry barrier — the five decisions that make an agent something a company can actually run.

  1. Online runtime and management: public and private nodes

    Running is built on online nodes, in two kinds. Public nodes need no environment of your own: billed by plan and token top-up, an employee is online as soon as it is activated. Private nodes run on your own deployment with your own model keys and infrastructure, so data and the inference path stay entirely under your control.

    Validate the business on a public node and move key employees to a private one — or go private from day one. Conversation, configuration, market and resource management feel identical; what changes is whose nodes they run on and how they are billed.

  2. Tools, MCP, skill documents and SOP customisable entirely online

    Model: choose the run node and the model. Resources: configure tools, MCP services, skill documents and execution workflows (SOP) online and map them to employees. Prompts: set prompts at employee level and session level, and on a shared employee a visitor prompt that applies to you alone.

    From using a ready-made employee to rewriting its capability for your own business, none of it requires leaving this online configuration. Customisation happens in the cloud workbench, not in local scripts.

  3. One tenant, user, role and authorisation system

    Employee runtime and tool or resource calls all go through the same tenant / user / role authorisation: employees and resources belong to the current project, with data isolation between tenants; who may create employees, call tools or use a given shared employee is decided by roles and grants, not by what the model feels like doing; sessions, usage, share review and accessible users are traceable under one identity system.

    That keeps agents running safely inside enterprise permission boundaries, instead of being an uncontrolled, open chat window.

  4. SaaS data — nothing is done to local data

    The data Xeromike-agent handles never touches your local disk or your local business database. Conversations, document references, resource configuration and runtime output all live in online object storage (S3-compatible buckets), isolated per tenant.

    Change device or browser and employees and sessions are still there; no local runtime to install, no local files to sync; the data boundary follows the tenant, not one particular computer.

  5. The lowest technical barrier

    For OPC / one-person companies and small teams we have prepared a large set of ready systems and tools, and we provide AI employees that are already configured. You do not need to write prompts, connect models or build MCP services: pick an employee in the market and start the conversation. When you need to go deeper, refine the prompts, map your own tools and SOPs, or deploy a private node.

    The goal is not to become an AI engineer first — it is to own digital employees that get work done.

Ways to use it

Four ways to put Xeromike-agent to work

Pick the path that matches how much control you need today; you can move between them later without changing tools.

  • Out of the box

    Who it suits teams that want someone working right now

    How to start Enter the employee market → use an employee that is already configured → start the conversation.

  • Light customisation

    Who it suits the flow roughly fits, but the wording has to be yours

    How to start After enabling a shared employee, write your own prompt; or name sessions and give a session its own prompt.

  • Fully self-built

    Who it suits clear business and tool requirements

    How to start Create your own employee → configure SOP / Skill / Tool / MCP → map them and go live.

  • Private deployment

    Who it suits you must use your own models and environment

    How to start Deploy a private node and bind your own model keys; employees are still managed in the same workbench.

Comparison

How it differs from an ordinary AI chat

The same conversation window on the surface; a very different system underneath.

A plain chat assistant compared with a Xeromike-agent AI employee
Dimension Ordinary chat assistant Xeromike-agent AI employee
Form One chat box Several manageable digital employees
Source of capability Mainly the prompt Model + tools + MCP + skills + SOP + prompt
How you get it Tuned up from zero by yourself Off the shelf from the market, or fully self-built
How it runs Vendor-managed, one way for everyone Public node pay-as-you-go, or a private node with your own keys
Data Often mixed with the local machine or a single account Tenant-isolated online object storage; local data is never touched
Security A weak or opaque permission model Unified tenant, user and role authentication

Both columns start from a conversation. The difference is what sits behind it: who the employee belongs to, which resources it may call, where it runs and where the data lives.

Get started

The goal is not to become an AI engineer first

It is to own digital employees that actually get work done. Start from the market, build your own employee, or run everything on a private node.

Open the employee market Create your own employee

  • Out of the box

    Enter the market, use an employee that is already configured, start the conversation.

  • Build your own

    Create an employee online and map SOP, Skill, Tool and MCP resources onto it.

  • Private node

    Deploy a private node with your own model keys — the same workbench on your own infrastructure. Talk to us.