Data inventory and canonical boundaries
One machine-readable catalog describes all 11 collections, their sources, schemas, APIs, measured counts, and the boundary of every projection. /data/latest/inventory.json ·/data/latest/coverage.json ·/data/latest/case-lift.json ·/data/latest/coverage-backlog.json
11
176
16
124
| Type | Count | Role | Source | API / schema |
|---|---|---|---|---|
observationObservation | 21 | A directly observable trace left by the hiring process. | data/{lang}/entities/observation/*.md | REST · schema |
barrierBarrier | 16 | A gate where the nominal hiring path can stop or diverge. | data/{lang}/entities/barrier/*.md | REST · schema |
mechanismMechanism | 28 | A causal force compatible with one or more observations and barriers. | data/{lang}/entities/mechanism/*.md | REST · schema |
patternPattern | 4 | A recurring, rule-defined combination of observations and compatible mechanisms. | data/{lang}/entities/pattern/*.md | REST · schema |
loopLoop | 3 | A validated reinforcing cycle among mechanisms. | data/{lang}/entities/loop/*.md | REST · schema |
interventionIntervention | 24 | An actor-owned change aimed at a barrier, mechanism, pattern, or loop. | data/{lang}/entities/intervention/*.md | REST · schema |
actorActor | 7 | A decision seat with explicit controls, blind spots, and incentives. | data/{lang}/entities/actor/*.md | REST · schema |
processProcess | 4 | A state machine naming the subject, states, owners, transitions, and expected timing. | data/{lang}/entities/process/*.md | REST · schema |
eraEra | 4 | A sourced period of hiring economics: capital, hiring behavior, entry paths, and its ending. | data/{lang}/entities/era/*.md | REST · schema |
evidenceEvidence record | 52 | A language-neutral source supporting registry claims. | data/evidence/*.md | REST · schema |
recordFinancial record | 13 | A durable budget, contract, funding, fee, payroll, subscription, or runway state with explicit flows. | data/{lang}/entities/record/*.md | REST · schema |
Validated scenarios · 16
Reusable situations composed from canonical entity IDs and explicit non-inferences.
Not for: A scenario is not an ontology entity or a claim that every case matches it.
Archetype overlays · 124
A nickname and grid placement for browsing every non-evidence ontology entry.
Not for: Archetypes are presentation, not evidence or canonical facts.
Formal substrate
A four-primitive projection used for proofs, conservation checks, and cross-cutting queries.
Not for: Never edit the substrate as a second source of truth.
Choose the tool by situation
Understand one concept in context
Open the reader page; follow related entries and evidence.
Website
Fetch one known canonical ID
GET the collection item, run hoba show, or call get_node.
Static REST APICLIMCP server
Find concepts without knowing an ID
Search all 11 ontology collections by ID, title, and prose.
CLIMCP server
Start from a recurring real-world situation
Use a validated scenario, then inspect its explicit non-inferences.
WebsiteCLIMCP server
Separate observations from compatible causes
Run the H→O→B→A diagnostic protocol; do not jump from signal to motive.
CLIMCP serverWebsite
Load the entire ontology into a data system
Use registry.json for one snapshot or NDJSON for streaming.
Registry JSONNDJSON
Analyze causal/topological relationships
Use graph JSON/GraphML/CSV; remember it is the six-type finding projection.
Graph exports
Build a typed integration
Generate a client from OpenAPI and validate payloads with JSON Schema.
OpenAPIJSON SchemasStatic REST API
Give an agent exact registry access
Connect the MCP server so the model calls tools instead of guessing from copied prose.
MCP server
Add or change registry data
Edit authored data only, run strict validation and the build, then review generated drift.
Authored sourceValidation tools
Verify formal invariants
Regenerate the formal projection and run the Lean checks.
Validation toolsLean projection
Static REST API
Read-only JSON, generated at build time and served straight from the CDN. There is no database behind it. /openapi.json
GET /api/v1/observations/index.jsonView JSON →GET /api/v1/barriers/index.jsonView JSON →GET /api/v1/mechanisms/index.jsonView JSON →GET /api/v1/patterns/index.jsonView JSON →GET /api/v1/loops/index.jsonView JSON →GET /api/v1/interventions/index.jsonView JSON →GET /api/v1/actors/index.jsonView JSON →GET /api/v1/processes/index.jsonView JSON →GET /api/v1/eras/index.jsonView JSON →GET /api/v1/evidence/index.jsonView JSON →GET /api/v1/records/index.jsonView JSON →GET /api/v1/graph/index.jsonView JSON →
Model Context Protocol (MCP) server stdio · local
Connect AI coding assistants and autonomous agents directly to the hoba knowledge graph.
// claude_desktop_config.json / .mcp.json — run from a local checkout
// (git clone … && pnpm install && pnpm build:packages)
{
"mcpServers": {
"hoba": {
"command": "node",
"args": ["/path/to/hoba.work/packages/mcp/dist/index.js", "--dir", "/path/to/hoba.work"]
}
}
}Available tools: get_registry_info, get_data_inventory, search_registry, get_node, get_scenario, analyze_social_complaint, find_nearby_scenarios, explain_observation, find_compatible_mechanisms, get_diagnostic_probes, find_patterns, get_interventions, traverse_graph, get_methodology, detect_temporal_anomalies, calculate_runway, verify_flow_conservation, evaluate_pattern_emptiness, validate_entity_ids, validate_scenario, validate_analysis, validate_claim.
Methodology resources: hoba://methodology/core, hoba://methodology/epistemic-rules, hoba://methodology/agency, hoba://methodology/evidence, hoba://methodology/non-goals.
MCP returns canonical data, rules, and deterministic checks. Natural-language reasoning can remain in the client model.
The packages are not on npm yet; npx @hoba/mcp will work once they are.
Command line interface (CLI)
pnpm build:packages && node packages/cli/dist/cli.js --helphoba search "reposted"- Find entities that match a query
hoba registry inventory --json- Get the canonical catalog, measured counts, surfaces, and usage guidance
hoba explain obs.materially_similar_role_reposted_shortly_after_rejection --stage technical- Run the hoba protocol on an observation
hoba scenario scenario.application_silence- Read a validated scenario composed over the ontology
hoba show mech.genuine_technical_skill_shortfall- Print one node in full
hoba validate --strict- Check content, the barrier DAG, loop declarations and English/Ukrainian parity
hoba latency proc.the_hiring_funnel_end_to_end recruiter-queue 45- Diagnose a stage dwell-time anomaly
hoba runway 25000 3500- Calculate the financial runway horizon of a search
Add --json to any command for machine-readable output.