About TW-Market Live Data Intelligence
TW-Market Live Data Intelligence is a local-first workbench for operators and AI assistants working with Taiwan-market evidence. It emphasizes identity validation, bounded and explicitly authorized execution, and separate audit and AI-ready handoff records rather than realtime trading.
Key features
- 01Identity-aware market requests
- 02Preview and authorize bounded execution
- 03Canonical results and separate audit packages
- 04Persistent revisioned Watchlists
- 05Official disclosure and revenue evidence
Use cases
- 01Validate a Taiwan-market target before a request
- 02Preview an operation before authorizing execution
- 03Review result provenance and its audit package
- 04Export evidence for AI-assisted discussion
MCP support
A separate stdio launcher exposes six governed tools for capability discovery, request validation and preview, result reading, AI-handoff export, and evidence retrieval. Request v3 is preferred, with v1 and v2 compatibility.
Configuration
Local server config is the same for every client. Copy it into your client’s MCP config file.
We haven’t collected a package or remote endpoint for this server. Follow the instructions in the project README.
Tools
Local servers expose tools only once running on your machine, and the maintainer has not provided a tool list.
Details
From the project READMETW-Market Live Data Intelligence
Local-first, governed Taiwan-market evidence for an operator and an AI assistant — validate first, preview bounded work, then explicitly authorize one execution where the capability permits it.

ProductVersion = 1.5.0. Selected v1.5.0 is unpublished; selection alone is not technical acceptance.
The current stable release is v1.0.0
(GitHub Release).
The published prerelease history is v1.0.0-rc.1
(GitHub Release),
and the earlier stable GitHub Release was v0.1.0. The current development
line now uses Unified Request/Result/Audit V3 as the preferred authority while
retaining V1/V2 compatibility. Phase G G1/G2 runtime capabilities have passed
bounded controlled-live acceptance; Phase H has one active TPEx attention route; broader
Roadmap Phase G/H scope remains explicitly incomplete. This is a local-first
evidence workbench, not a realtime trading product.
Why TW-Market
AI discussion needs evidence with identity, source, timestamp, caveat, and execution provenance — not an unqualified price claim. This workbench gives a human operator a governed local path from request validation to an AI-ready handoff, while retaining a separate audit package.
Quick start
git clone https://github.com/SmartyJohnway/tw-market-live-data-intelligence.git
cd tw-market-live-data-intelligence
python -m venv .venv
# Activate .venv using your shell, then:
python -m pip install -r requirements-lock.txt
python scripts/verify_environment.py
python scripts/manage_security_master.py status
NOT_INITIALIZED is the normal fresh-install state. A Security Master update
is an explicit operator action and may use official external acquisition:
python scripts/manage_security_master.py update --live
python scripts/run_unified_workbench.py
Open the loopback Workbench at /workbench/.
For an MCP host, start the separate stdio launcher:
python scripts/run_unified_market_evidence_mcp.py
A governed 2330 workflow
- Create or select an installation-local Watchlist, then add
2330. - Compose a Unified Market Evidence Request and validate its identity.
- Preview the planned operation and its capability boundary.
- Explicitly authorize and confirm one bounded execution only when the preview is executable.
- Read the canonical Result and Audit Package, or export the AI-ready handoff for continued discussion.
Core capabilities
- Identity-aware requests: Mode A validates targets against the installation-local Taiwan Market Identity Service.
- Bounded execution: Mode B previews, explicitly authorizes, and executes one request only when its governed capability is executable.
- Auditable handoff: Mode C creates a canonical Result, separate Audit Package, and AI-ready Markdown without dispatching another market source.
- Persistent Watchlists are supported: installation-local Watchlists have immutable revisions, optimistic concurrency, and explicit preview/commit mutation.
- Official research evidence: eligible TWSE/TPEX common shares support the latest completed official material-disclosure batch and latest available monthly revenue (TWD, unit thousand), using official CSV with an explicit governed official JSON fallback.
Unified MCP
The MCP surface has exactly six governed tools:
market_describe_capabilities, market_validate_request,
market_preview_request, market_read_result,
market_export_ai_handoff, and market_fetch_evidence.
Request v3 is preferred; Request v1 and v2 remain accepted compatibility contracts. New governed materializations default to Result v3 and Audit Package v3, while historical V1/V2/V3 packages remain readable without rewriting. See the V1 historical public contracts and current AI usage guide for request, result, and handoff semantics.
MCP distribution
The supported Windows MCPB distribution is published with
v1.0.0:
download the MCPB
and its SHA-256 sidecar.
It uses local stdio transport and provides the same six tools. The official
Registry entry
is io.github.SmartyJohnway/tw-market-live-data-intelligence. See
MCP distribution for installation
scope, integrity, and Security Master boundaries.
Data and source caveats
Capability support, currentness, source provenance, and execution eligibility
are explicit product data. A supported identity does not imply an executable
source route; a preview is not an authorization; a source observation is not a
realtime guarantee. Research support does not provide arbitrary disclosure or
monthly-revenue history, realtime disclosures, or financial_summary.
Consult the capability matrix,
source matrix, and
governance boundaries before relying
on any result.
Safety and non-goals
There is no automatic polling, scheduler, startup market fetch, Watchlist-driven automatic execution, trading, order routing, full-market scan, model-selected URL/executor, or realtime guarantee. Persistent Watchlists mutate only through explicit preview/commit. Never commit credentials, tokens, cookies, or private market payloads.
Documentation
Repository governance front door:
- PROJECT.md — current project/product authority
- ROADMAP.md — canonical Roadmap V3.2
- HANDOFF.md — current operational handoff
- AGENTS.md — AI/repository collaboration rules
Product documentation:
Contributing and security
Read CONTRIBUTING.md before opening a pull request. Report security concerns using the grounded guidance in SECURITY.md.
Release status
VERSION records 1.5.0 (selected, unpublished). The current stable release is v1.0.0
(GitHub Release);
published prerelease history remains v1.0.0-rc.1. The current development
line has progressed through accepted Phase G G1/G2 runtime work and H-ACT-V3
preferred-authority promotion; this does not retag or rewrite the immutable
v1.0.0 release. Roadmap Phase G/H completion remains a broader scope question
tracked separately in ROADMAP.md. Both release tags are immutable authorities; subsequent
documentation status updates do not move them.
Project Overview
This root README is the current product entry point. Engineering history,
protocol acceptance, and prior M5/M6/M8 architecture remain available through
Project History, the Engineering history / protocol
archive, and
docs/archive/, including the
2026-06-30 historical README.
They are retained for audit and compatibility,
not as a second current-product contract.
License
See LICENSE.
Comments
New to MCP?
From what the protocol is to installing your first server, these answers cover most first questions.
An open protocol that defines how AI applications connect to external tools and data in a uniform way. Clients talk to the model; MCP servers provide tools, resources and prompt templates. They exchange standard messages, so one server can be reused by many clients.