Immutable LDS schemas, verified semantic alignments, and an MCP gate that forces enterprise agents through deterministic governance before they touch state.
MCP stdio by default. Use etg serve --transport sse for HTTP/SSE.
External LLMs are treated as untrusted callers. They can request schemas, propose alignments, and log conflicts, but every action must pass the Entigram gate before it reaches the ledger.
Unknown entities, invented attributes, and unapproved joins are rejected against explicit LDS contracts. The agent gets a structured error instead of permission to drift.
Run Entigram as governance infrastructure in front of Claude, Gemini, LangGraph, or internal agents. The server exposes deterministic tools for schemas, alignments, and conflict logging.
# Starting zero-trust control plane...
etg serve
# Agent requests governed schema boundaries...
tool etg_get_schemas
# Proposed mappings are validated before ledger write...
tool etg_propose_alignment
OK LDS schema boundary loaded
OK invalid mappings rejected at the gate
Zero-trust governance active.
Explicit LDS contracts define the entities and fields agents may reference. Anything outside the schema is rejected before state changes.
Semantic mappings are proposed through MCP tools, validated against schema boundaries, and stored with provenance in the SQLite ledger.
Conflicts, decisions, and alignments are persisted locally with WAL-mode concurrency so every governed action has an audit trail.
Publish ai-catalog.json so CLI agents and enterprise orchestrators can discover Entigram and route work through the semantic gate.