wquestions-mcp
The wquestions-mcp server lets you build and query a structured knowledge model of any domain using a fixed 7-axis framework (who, what, where, when, how much, which kind, how) — no custom schema required.
Explore the framework — Use
list_axesto understand the 7 modeling axes andlist_rolesto see canonical typed role connectors.Build a knowledge model — Add named entities to any axis (people, places, objects, times, magnitudes, categories) with
add_entity, and optionally pre-register situation/verb types withdefine_verb(or let the server auto-register on first use).Assert facts — Use
assert_situationto record any real-world event or relationship by attaching role-labeled arguments; supports time-bounded facts viavalid_from/valid_to.Query the model — Use
askto fix some roles and retrieve others, filter by situation type, or query the model as it existed at a specific point in time.Correct/update facts — Use
correctto update a role on an existing situation in an append-only manner, preserving full history.Inspect the universe — Use
show_modelto dump all entities and facts, including persistence metadata.Try instant demos — Use
load_example(e.g. the "spa" example) to load a prebuilt domain and start querying immediately.Reset and start over — Use
resetto clear the current universe and begin modeling a new domain.Persist across sessions — The universe is saved to an append-only log (
~/.wquestions/universe.jsonlby default) and automatically reloaded on restart; configurable via theWQUESTIONS_LOGenv variable or disableable withWQUESTIONS_LOG=off.
Click on "Deploy Server".
Wait a few minutes for the server to deploy. Once ready, it will show a "Started" state.
In the chat, type
@followed by the MCP server name and your instructions, e.g., "@wquestions-mcp Load the spa example and show me the model."
That's it! The server will respond to your query, and you can continue using it as needed.
Here is a step-by-step guide with screenshots.
wquestions-mcp
Model any domain in 7 questions.
An MCP server that lets Claude Desktop (or any MCP client) build and query a knowledge model of anything — a spa, a barbershop, a clinic, a bank — using one fixed set of tools. No per-domain schema to write, ever.
The problem
Every domain today gets its own bespoke ontology or database schema: a CRM schema for sales, a clinical model for a clinic, a different one again for a bank or a taxi dispatcher. None of it transfers between domains, and none of it was designed for an LLM to reason over — each new domain means new modeling work before an AI can even start answering questions about it.
WQuestions replaces all of that with a single fixed index: 7 axes that any fact, in any domain, answers. Model a domain by asserting facts on those 7 axes; query it the same way no matter what the domain is.
Related MCP server: querywise-mcp
The 7 axes
Axis | Question | Holds |
Q | who | agents |
O | what | objects, and reified situations (facts treated as things) |
L | where | places |
T | when | time points and intervals |
N | how much | magnitudes with a unit |
K | which / what kind | atemporal categories, types, states |
M | how | the predicates that connect Q/O/L/T/N/K to each other — structural, not a value axis |

Quickstart
Add this to your Claude Desktop config (claude_desktop_config.json) and
restart Claude Desktop:
{
"mcpServers": {
"wquestions": {
"command": "uvx",
"args": ["wquestions-mcp"]
}
}
}Prefer to run from source? Clone this repo and, from the repo root,
pip install -e . into a virtualenv (the engine is bundled — no other
package needed). Then point command/args at that venv's wquestions-mcp
script (e.g. command: ".../.venv/bin/wquestions-mcp", args: []) instead
of uvx.
Then ask Claude: "Load the spa example, then show me the model." See
DEMO.md for the full 30-second walkthrough.
Tools
Tool | Does |
| Describe the 7 axes |
| List canonical roles (who/what/where/... connectors), typed by domain and range |
| Create an individual on a value axis (Q, O, L, T, N, K); for the N axis pass |
| Register a situation type and its roles — optional, |
| Assert a fact: reify a situation and attach its roles |
| Correct/update a role on an existing situation by re-asserting it (append-only; |
| Query by projection — fix some roles, ask for others, optionally as of a point in time; pass |
| Dump the current universe: every entity and fact |
| Load a prebuilt demo universe ( |
| Clear the model and start a fresh universe |
How it works
The LLM client does the natural-language-to-structure step: it reads "Diego
cut Marco's hair at Barber Kings on 2025-06-11" and turns it into
role-labeled arguments (agente: diego, paciente: marco, lugar_de: barber_kings). The server never parses English — it takes those roles,
validates them against the 7-axis model, and runs ingest and query over the
wq engine. Same engine, same 10 tools, whether the domain behind them is a
spa or a barbershop.
Persistence
By default the server persists your universe to an append-only log and reloads on restart, so a modeled domain survives across sessions — you never rebuild by hand. Nothing is ever overwritten; corrections are appended and the log is the source of truth.
Default location:
~/.wquestions/universe.jsonl.Choose a file: set
WQUESTIONS_LOG=/path/to/domain.jsonl. Point each domain at its own file (e.g. one per MCP server entry in your client config) — a single log holds one live universe at a time (aresetstarts a fresh domain within it).Turn it off (pure in-memory):
WQUESTIONS_LOG=off.
show_model reports the active log path and how many events were replayed, so
you can confirm persistence is on.
Further reading
The 7-axis model — why it's fixed, what each axis actually commits to, and how it holds up once you push on it — is worked out in full in WQuestions, the book this project comes from (Spanish). Start with Chapter 8, El espacio multidimensional for the axis model itself, or the table of contents.
Available Tools
9 toolsadd_entityC
Create an individual on a value axis (Q, O, L, T, N, or K).
Inputs: axis (required), entity_id (required), label. Returns: result
askC
Query by projection: fix some roles, ask for others. type filters to
situations whose type matches the given category id (auto-registered
verbs get the id action_<verb>). at (ISO) queries the model as it
was valid at that time.
Inputs: at, ask, type, fixed. Returns: result
assert_situationA
Assert a fact: reify a situation for verb and attach its roles.
Each role value is an existing entity id or an inline {id, axis, label}.
valid_from/valid_to are ISO-8601, for facts that are only valid during
a time range.
Inputs: verb (required), roles (required), extra, valid_to, +1 more. Returns: result
define_verbA
Register a situation type (verb) and which roles it takes. Optional: assert_situation auto-registers unknown verbs permissively.
Inputs: verb (required), situation_type (required), optional, obligatory. Returns: result
list_axesA
List the 7 WQuestions axes (Q who, O what, L where, T when, N how-much, K which, M how) and what each is for.
Inputs: no named parameters. Returns: result
list_rolesA
List the canonical roles with their typed signatures. Unknown roles are also allowed (liberal policy) — you may invent domain roles freely.
Inputs: no named parameters. Returns: result
load_exampleB
Load a prebuilt demo universe (e.g. "spa") to try queries instantly.
Inputs: name (required). Returns: result
resetA
Clear the model and start a fresh empty universe.
Inputs: no named parameters. Returns: result
show_modelA
Dump the current universe: entity/fact counts and every fact.
Inputs: no named parameters. Returns: result
Tool Schema Changelog
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
Loading tool schema changelog…
TDQS
Scored across 9 tools
Each tool targets a distinct operation: adding entities, querying, asserting situations, defining verbs, listing axes/roles, loading examples, resetting, and showing the model. No overlap in functionality.
Most tools follow a verb_noun pattern (add_entity, assert_situation, define_verb, list_axes, etc.), but 'ask' and 'reset' are single verbs, introducing minor inconsistency.
9 tools is well-suited for a knowledge graph modeling server, covering core operations without being excessive or sparse.
Missing operations like updating or deleting specific entities or situations, and lacking targeted retrieval (e.g., get_entity by id) limit the completeness for a full lifecycle.
Maintenance
Related MCP Connectors
Agentic CRM for service businesses — bookings, customers, WhatsApp, loyalty, invoicing.
Verified local businesses, bookable by AI agents: services, prices, availability and appointments.
Run a local business from your AI app: bookings, customers, inbox, website, posts and payments.
Voice agents that book into the software your clients actually run, built and audited from chat.
Related MCP Servers
- AlicenseAqualityAmaintenanceAgent-native semantic layer, letting AI agents query databases through specifying intent instead of writing SQL, then compiling structured queries into correct, dialect-aware SQL. Dynamic and expressive, supporting multi-stage queries, time-shifts, and complex join schemas.2012,719 PyPI348MIT
- AlicenseAqualityCmaintenanceQuery SQL databases (SQLite, PostgreSQL, BigQuery, Databricks) in natural language through a business semantic layer — glossary, metrics, and a data dictionary grounded against your real schema. Read-only by default, with an embedded SQLite + sqlite-vec metadata store and no external infra required.253MIT

EnrichMCPofficial
AlicenseNot gradedqualityDmaintenanceTurns your data model into a semantic layer for AI agents, automatically generating typed, discoverable tools with entity relationships and schema discovery.26 PyPI644Apache 2.0- AlicenseNot gradedqualityDmaintenanceEnables natural language querying of databases with multi-turn conversations, auto-generated charts, and proactive monitoring via scheduled queries and alerts.1MIT