v0.8.3 · The Organizational Canon Layer

The authoritative truth for your AI workflows.

Don't let your AI assistants improvise. MsgStack structures compliance manuals, product specifications, and brand positioning into a version-controlled, queryable knowledge graph.

View on GitHub The Canon Model →

Structured Canon Sample

canon_spec.md

# Canon Domain: Developer Specs
## Entry: API Limits
Owner: Engineering Lead
Rule: Rate limit cap at 100 req/sec.
Status [LOCKED]: "Active on all /v2/ paths."

# Canon Domain: Message House
## Entry: Brand Pillars
Owner: Product Marketing
Rule: "98% of zero-day threats stopped proactively."
Status [LOCKED]

markdownUTF-8

Approximate RAG vs. Deterministic Traversal

Standard vector databases approximate answers based on nearest-neighbor semantics. MsgStack structures exact boundaries, guaranteeing that version-locked rules are never paraphrased.

Legacy Vector Approximation

Vector databases convert your PDF policies into float arrays. During a query, an LLM retrieves similar paragraphs and attempts to summarize. Under pressure, the model paraphrases compliance rules, invents pricing metrics, and hallucinated features.

vector_rag_metrics.logRAG

Retrieval: Approximate KNN Search
Accuracy: Probabilistic (85% - 95%)
Downstream Drift: High

RECOMMENDED

MsgStack Canon Grounding

MsgStack parses documentation into deterministic adjacency tables. When an agent queries, it retrieves locked claims verbatim. Explicit dependency links ensure that if a root policy is updated, all downstream templates flag for review.

canon_traversal_metrics.logCANON

Retrieval: Graph Path Traversal + Locked Match
Accuracy: 100% Deterministic
Downstream Drift: Controlled / Versioned

FLAGSHIP DOMAIN SCHEMA

Bring positioning control to GTM AI with Message Houses

The "Message House" is our flagship Canon domain template. Instead of letting sales and marketing assistants prompt LLMs with generic guidelines, MsgStack stores brand claims, value pillars, and audience personas in structured adjacency nodes.

When Claude or Cursor drafts an email sequence or battlecard, it queries the message house schema directly—ensuring all generated materials cite exact, legally approved claims.

Explore the Message House Schema →

DOMAIN TYPE: MESSAGE_HOUSEACTIVE

Pillar 1: Autonomous Risk Neutralization
  ↳ Persona Target: CISO, Director of IT Sec
  ↳ Locked Claim: "SOC2 Compliance certified across all European data sectors."

Pillar 2: Zero-Latency Operations
  ↳ Persona Target: Head of Infrastructure
  ↳ Locked Claim: "Sub-millisecond query verification using local SQLite caching."

Grounding In Action

Interactive Assertion Verification

MsgStack runs a validation engine to check if AI outputs align with approved Canon assertions. Run the grounding checker below to simulate scanning a generated SDR email for ungrounded claims.

AI Draft InputTarget: outbound-email-sdr

Nexus Cloud protects your assets. With our system, you get 98% of zero-day threats stopped proactively. Our infrastructure provides sub-millisecond query verification and we back our product with a 99.999% uptime SLA guarantee to ensure operations run without interruption.

Verification Audit

100%

ALIGNMENT SCOREPending Verification

Claim Inspector

Click Run Grounding Checker above to verify the text.

Active Domain Contexts: developer-specs, security-spec, corporate-narrative

In Practice

Four tool calls. Fully grounded output.

Once connected, your AI assistant queries your canon layer directly. Here's a real session grounding a sales email against the PMM Message House.

Discover

list_canon_domains returns every approved truth framework your team has published — with entry counts and lock status.

Retrieve

search_canon pulls version-locked claims relevant to your task. Exact matches, never paraphrased.

Generate

generate_artifact produces output cited directly against the active domain. Every claim has a source entry.

Audit

score_alignment catches ungrounded claims before they ship — flags anything without a matching locked entry.

claude_mcp_session.logLIVE

> list_canon_domains

  ● PMM Message House     [ACTIVE] 12 entries · 8 locked
  ● Engineering Specs     [ACTIVE] 34 entries · 29 locked
  ● Legal SOC2 Policies   [STAGED]  7 entries · 2 locked

> search_canon "zero-day detection rate"

  Found: Entry 441-B [LOCKED]
  Claim: "98% of zero-day threats stopped proactively."
  Owner: Product Marketing · Verified: 2024-11-01

> generate_artifact type=sales_email domain=pmm_house

  Grounding against 4 locked entries...
  Draft generated. Alignment: 100%
  All claims cite approved canon entries.

> score_alignment "Our platform stops threats instantly"

  ⚠ Claim: "instantly" — NO MATCH
    Nearest: "sub-millisecond" [Entry 228-C, DRAFT]
  Score: 71% · 1 claim flagged for review

Governed Canon Dashboard

View alignment scores, configure domain schemas, verify graphs, and track drift in GTM/Product documentation.

mcp.abidc.dev

Centralized Canon Domains

Manage specs, policies, and positioning across every department.

mcp.abidc.dev

Grounding & Alignment Metrics

Track downstream drift and verify which facts are locked vs. pending review.

Hosted SaaS Waitlist

MsgStack is open source. Want us to host and manage it for you? Join the waitlist for the managed SaaS version.