tdcv2-mcp
v0.1.6
Published
MCP server for TDCv2, a fake and test data generator: agents write a .tdc config and generate realistic mock, synthetic and test data locally — CSV, JSON, SQL; same config, same bytes.
Maintainers
Readme
tdcv2-mcp
An MCP server that lets an AI agent generate fake, mock, synthetic and test data with
TDCv2:
the agent writes a small .tdc config, the engine generates the rows on your machine,
and the same config gives the same file byte for byte on every run. Nothing is sent
anywhere — the server only runs the engine locally.
It is the same knowledge as the TDCv2 skill, as tools — for clients without a terminal, or
that prefer tools to skills. The skill itself ships in this package:
npx -y [email protected] install-skill puts it where your agent looks for skills (--help for
the options).
Connect it
Claude Code:
claude mcp add tdcv2 -- npx -y [email protected]Claude Desktop, Cursor and other clients — the usual mcpServers entry:
{ "mcpServers": { "tdcv2": { "command": "npx", "args": ["-y", "[email protected]"] } } }Windows: use "command": "npx.cmd" if the client cannot start npx.
The server works in the folder the client starts it in — the project. It reads .tdc
files and writes generated files only inside that folder. tdc_pack_add puts packs where
the project's TDCv2 config says (by default tdcv2-packs/ in the project).
Tools
| tool | what it does |
| :--- | :--- |
| tdc_read_docs | the guide (read once), traps check does not catch, test code in five languages, several linked tables in one file, any reference page |
| tdc_find_packs | data packs installed here, by words — the paths for <gen type="template"> |
| tdc_pack_add | install packs for another language or country (fr, france) — tdcv2 pack add, after tdcv2 init in a project without a pack store |
| tdc_check | every diagnostic with code, line, column, hint and "did you mean" |
| tdc_generate | preview (at most 50 lines) or write the file; a failed <assert> comes back as row, message, condition, values |
| tdc_peek | a short summary of a CSV, JSON or SQL file — rows, splits, ranges; a JSON object of arrays table by table, with every <x>_id checked against <x>s.id |
| tdc_format | pretty-print a config like tdcv2 format |
Also a prompt, tdcv2_guide, and resources tdcv2://guide, tdcv2://traps,
tdcv2://from-code, tdcv2://linked-tables.
Which engine
The project's own tdcv2 if it has one (npm i -D [email protected]), else the server's own
dependency — the version this release is checked with. TDCV2_ENGINE=<package folder>
pins one.
Checked
test/smoke.mjs drives the server through a real MCP client over stdio — every tool, the
prompt, a resource — and checks that the file it writes is byte for byte the one the CLI
writes. CI runs it on Linux and Windows.
