Grimoire

The hidden architecture of corporate intelligence.

A secure aggregation core for business knowledge: connecting multiple databases and source systems into governed, AI-safe intelligence without blocking how people build, change, consume, or create.

Connect Products, services, capabilities, evidence, SLAs, vendors, case studies and approved text.

Govern Secure enclave, provenance, approval state, policy scope, masking and AI eligibility.

Consume Human workflows, process automation, proposal systems, reporting and AI agents.

Built to feed the surfaces where offer knowledge fails first

BidFoundry approved bid language

Lead qualification fit and evidence signals

Architecture agents patterns and dependencies

Service design SLA and boundary rules

Reporting consistent offer metrics

Proposal language reusable approved text

Governed knowledge flow.

Grimoire keeps the authority in the canonical core, then publishes only approved, policy-scoped projections to people, tools, and agents.

Source inputs SharePoint, CSV, SMEs, service docs

Grimoire core stable IDs, provenance, approval state

Policy layer AI eligibility, masking, role and purpose

Controlled outputs BidFoundry, agents, reporting, proposals

One semantic model. Many controlled projections.

Grimoire harmonises source systems into approved, versioned records and emits scoped responses with record IDs, owners, approval state, effective dates, and evidence.

Canonical envelope

stable_idcommon reference across tools
owner_stewardaccountability before reuse
approval_statedraft content cannot feed bots
effective_datesstale claims fail closed
evidence_linksclaims carry proof

Editors + SMEs

Workflow + validation

Canonical core Sensitive annex Object storage Audit ledger

Policy + query composition

Search Graph Vector API

Approval coverage94%

Status	Evidence	Risk	Owner	Action
Approved	SLA matrix	Low	Service	Use
Review	Vendor claim	Medium	Partner	Hold
Draft	Pattern note	Blocked	Architecture	Mask

Not a bigger file store.

Grimoire keeps source systems in place and governs the knowledge layer between them and consuming tools. Search, vectors, caches, proposal contexts, and analytics views are derived read models, not authorities.

Stable identities

Products, services, capabilities, vendors, SLAs, evidence, and text blocks carry persistent IDs across consuming systems.

Provenance by default

Every answer can point back to source URI, version, owner, effective date, approval state, and evidence record.

Purpose-scoped access

Human and bot clients receive only the fields and classifications their role, purpose, and market context allow.

Future development phases.

The current plan keeps Grimoire evidence-led: prove the governed model, then expand the product surface and only then commit to the longer-term platform shape.

Phase 01

Domain taxonomy and approval model

Define the core entity families, ownership rules, approval states, maturity signals, classification, and AI eligibility gates.

Month 1

Phase 02

Thin-slice canonical schema

Stand up the PostgreSQL canonical model for capabilities, products, services, vendors, evidence, approved text, and provenance.

Months 1-2

Phase 03

MVP control plane

Use Directus over PostgreSQL for early admin/editor acceleration, backed by deterministic imports, search, API, MCP, and policy filters.

Months 2-3

Phase 04

Bot-safe query and provenance

Return scoped responses with record IDs, versions, approval state, effective dates, provenance, masking, and AI eligibility checks.

Months 3-4

Phase 05

Product front end and resilience tests

Build the Grimoire-native editor, review queue, source explorer, bot preview, restore, redaction, and explainability checks.

Months 4-5

Phase 06

Platform decision and scale plan

Decide whether Directus remains back-office, is removed, or Grimoire moves toward a fuller custom product platform.

Months 5-6