bolor.
TycLocal work & business operations

Tyc / Conversational marketplace / Mongolia

A local marketplace,
managed through
conversation.

Bolor character, used in Tyc’s chat brandingTycVisit the product ↗

Connecting people who need help, independent gig workers, and local businesses—with Bolor chat as a shared entry point for finding work and getting things done.

Product
Tyc · tyc.mn
Market
Local work & services in Mongolia
My contribution
Conversational work using my own Bolor Brain MCP server
Platform
Responsive web application
Bolor · Current conversation entry Product screen
Current Tyc public chat entry screen in Mongolian, with Bolor greeting, a composer, role selection, and suggested starting prompts.
Current public chat entry, captured from Tyc on October 7, 2026. Suggested prompts and role selection give people a place to start.

01 / The challenge

One marketplace.
Several ways to participate.

A customer may need a cleaner. An independent worker may be looking for a short job. A service business needs to fill its schedule and manage the customers it already has.

Tyc connects these needs within one ecosystem. The design challenge is to make the next task understandable without asking people to learn every part of that ecosystem first.

The central interaction: express an intention in chat, review a structured result, then take the next action.

Current Tyc marketplace landing page connecting customers, independent workers, and businesses.
Current marketplace landing page, connecting customers, independent workers, and businesses.

02 / Three perspectives

The same platform.
A different starting point.

Explore the customer, worker, and business paths. Each view follows functionality present in the project.

Customer

Turn a need into a
local connection.

A request gathers the service, location, budget, and timing. People can compare replies from an individual or a business before continuing the conversation.

  1. Describe the taskCreate a request draft through Bolor.
  2. Review & publishCheck the details and confirm the request.
  3. Compare responsesSee provider identity, price, ratings, and context.
  4. Continue with a personAccept a response and move into a request-linked conversation.
Customer illustration on the current Tyc website comparing individual and business cleaning responses.
Current live-site customer illustration—not an authenticated request screen. Prices and replies are illustrative.

My own conversational tooling

Bolor Brain MCP.
A system I built and used.

I used my own Bolor Brain MCP server for the conversational work behind Tyc. Building the supporting AI tooling was part of my contribution, alongside shaping the marketplace experience.

Bolor Brain provides structured reasoning, persistent memory, and learning from previous problem–solution cases. Its public implementation includes several reasoning approaches and tools for storing and recalling knowledge.

Tyc’s application layer carries out marketplace actions and returns their results as cards in the conversation. This keeps reasoning, product actions, and the user’s confirmation steps distinct.

Explore my Bolor Brain MCP on GitHub ↗

Reasoning

Choose an approach

Symbolic, knowledge-graph, case-based, hypothesis, and analogical reasoning are coordinated through a hybrid reasoning tool.

Memory

Carry knowledge forward

Store and recall facts, cases, and relationships through the server’s memory tools.

Learning

Retain the experience

Save a problem, its solution, and its outcome so future reasoning can draw on previous work.

03 / From words to action

Make the intention
visible before acting.

Free-form input is flexible. Structured cards give people something concrete to inspect, compare, and confirm.

Current Tyc public mobile chat entry with Bolor greeting, suggested prompts, and composer.
Current mobile chat entry. The accompanying request lifecycle is traced from the application code.
01

A draft is a separate state

The service-request flow saves a draft before publication. Category, location, budget, and timing become visible fields, giving the conversation a concrete object to work with.

02

Publishing is an explicit step

Confirmation checks that the draft belongs to the current person and is still a draft. A phone number is required before publication; the interface points to settings if it is missing.

03

Matching carries local context

Provider matching uses service category and location, with ratings and distance contributing to ranking when available. Both independent providers and businesses can appear.

04

The handoff keeps the task attached

Accepting a response creates or retrieves a conversation associated with that request. The customer and provider can continue discussing the same work.

IntentionDraftConfirmationProvider repliesDirect conversation

Keeping context

AI starts the task.
People agree on the work.

Bolor’s assistant conversation and the customer–provider message thread serve different purposes. The assistant helps navigate and act; direct messaging lets the people doing the work settle the details.

The current gig feed starts a conversation when a provider expresses interest. Accepted responses in the service-request flow also retain the request link. The diagram below summarizes this implemented handoff.

CustomerPublish a request

Service, location, budget

ProviderExpress interest or respond

Independent worker or business

Shared contextContinue the conversation

The request stays attached

Flow diagram based on the current application code.

04 / Business operations

Chat as the front door
to a working business.

Business accounts need more than incoming leads. Tyc’s chat tools cover bookings, clients, services, revenue, and business settings, while dedicated screens provide an overview.

Setup follows a guided conversation with focused inputs for business type, name, services, prices, and opening hours. Structured widgets keep these details reviewable.

Business operations illustration from the current Tyc website with example booking, revenue, and customer summaries.
Current live-site illustration of business operations; it is not a capture of an authenticated dashboard.

Context

Capabilities follow the account

Chat distinguishes customer actions from business actions using the signed-in account’s business context.

Local relevance

Mongolian by design

Mongolian interface copy, tögrög pricing, and location-aware discovery make the marketplace specific to its audience.

Continuity

Actions return visible results

Tool results appear as cards inside the conversation, connecting the user’s request to the resulting product state.

05 / What comes next

Validate the whole journey,
from the first message
to the agreed task.

What is present

Tyc has a public website and a codebase covering conversational requests, provider matching, direct messaging, guided setup, and business management. This case study traces those implemented flows.

What needs validation

Can people correct a misunderstood request? Do provider comparisons support confident choices? Does the handoff preserve enough context to reach an agreement? These are the next usability questions—not measured results.

Visuals were captured from the current Tyc website on October 7, 2026 and checked against its source components. Chat captures show the public entry state. Marketplace and business illustrations come from the live landing page and contain illustrative records, not authenticated account data. This case study does not claim measured adoption, conversion, or revenue impact.