Monogram Language

Monogram Language is the language-evidence laboratory beside Monogram. It studies compiler, runtime, standard library, package-manager, validator, grammar, and ecosystem-control repositories to derive lifecycle-role and language-semantic profile evidence. It does not change Monogram ranking. Every promotion, score, connector, and live-analysis surface keeps ranking_ready=false until the corresponding candidate contribution can be shown in Monogram score-debug and evaluated without benchmark-answer leakage. Published contract: monogram-language 0.1.3, OpenCLIs build 2026-07-12-192026. The executing binary, OpenCLIs registry, and public CDN metadata agree on this release. The CLI exposes 18 operational root commands plus the machine-readable mcp-schema / --mcp-schema surface. Root help, --help, and -h print the full embedded guide. Subcommands do not implement their own --help; for example, capabilities --help exits 1. version, --version, and -V are identity forms rather than operational commands.

Quick Start

openclis install monogram-language

monogram-language --version
monogram-language registry --lang rust
monogram-language analyze --lang rust --repo /path/to/rust \
  --corpus-id rust:runtime --domain rust-runtime --kind runtime \
  --json --out rust-runtime.json
monogram-language validate --profile rust-runtime.json

Complete Command Reference

Command	What it does	Persistent effect or key controls
`analyze`	Scans a language-self corpus and emits identifier roles, layout/structure, syntax-shape, parsed AST, grammar metadata, semantic ontology/flow, transfer patterns, source/leakage gates, and optional semantics-core evidence.	Requires `--lang` and `--repo`; use `--corpus-id`, `--domain`, `--kind`, `--json`, `--out`; `--semantics-core*` enables the gated core lane.
`capabilities`	Prints active, planned, and gated capabilities plus their evidence, ranking, score-debug, and Monobench boundaries.	`--json`; read-only.
`coverage-gap`	Audits every Monogram-supported language for corpus plans, grammar availability, profile artifacts, and semantics-core routing coverage.	Repeat `--profile`, `--profile-dir`, or `--profile-manifest`; `--json`; read-only.
`objective-matrix`	Prints the whole lab’s objective/status checklist, missing proof, and next experiments.	`--json`; artifact-only.
`score-debug-plan`	Lists the exact future `language_profile.*` and candidate-local fields required before ranking integration.	`--json`; artifact-only.
`semantic-ir`	Converts aggregate Monogram Semantics stats into a gated semantic-IR bridge review.	`--stats <json>`, optional `--source-version`, `--json`; never imports raw function names.
`semantics-core`	Runs `lib-monogram-semantics-core` directly and produces a language-owned absorption review.	`--project`/`--path`, `--db`, `--compiler-fact`, `--lifecycle-fn`, `--json`; may update the semantic DB, never a language profile or ranking.
`induce-lexicon`	Learns aggregate token distributions from Semantics-provided structural labels.	`--anchors <json>`, optional `--corpus-profile`, `--source-version`, `--json`; raw names are not retained.
`connector`	Validates and summarizes an AI-readable CLI-to-CLI connector contract without executing its phases.	`--file <connector.json>`, `--json`; read-only review.
`registry`	Lists recommended language-self corpora by language and corpus kind.	`--lang`, `--json`; read-only.
`validate`	Checks a saved profile and mirrors its evidence-source, leakage, ontology, and benchmark gates.	`--profile <file>`, `--json`; does not rewrite the profile.
`compare`	Compares two or more validated profiles, including role rates, surface overlap, AST/grammar/flow similarity, drift, and shared-domain-port risk.	Repeat `--profile`; `--json`; artifact-only.
`delta`	Compares two profile snapshots and emits bounded current-state drift evidence.	`--before`, `--after`, `--json`; requires matching ontology/corpus identity for live-ready evidence.
`incremental`	Re-analyzes a current repository against an accepted baseline and emits the same profile-delta contract.	`--baseline`, `--repo`; `--out-current` persists the refreshed profile, `--out` persists the report.
`loop`	Builds a repeatable research report over profiles, queries, and optional incremental/semantic/lexicon/coverage reviews.	Repeat `--profile`; optional `--query` and review-report flags; `--out`, `--json`.
`promote`	Produces a promotion-review artifact for corpus-confirmed roles, terms, transfer patterns, and blocked semantic-sitter candidates.	Repeat `--profile`; optional review reports; `--out`, `--json`; never edits ranking.
`score`	Scores one natural-language query against one saved profile and explains all gates.	`--profile`, `--query`, `--json`; evidence only, not a root-cause verdict.
`seed`	Shows default source-extension seeds for a language key.	Optional `--lang`; read-only.
`--mcp-schema`	Prints the typed CLI schema used for MCP wrapping.	Machine-readable; no analysis run.

The live schema currently contains the same 18 operational commands. It is a description surface only; printing it does not start an MCP server or run a profile analysis.

Evidence Flow

language-self repositories
        ↓ analyze + validate
saved corpus profiles
        ↓ compare / delta / incremental
review evidence
        ↓ loop / promote / score
explicit score-debug requirements
        ↓ separate human-approved Monogram integration
Monobench canary evaluation

Persistence and Review Boundaries

analyze --out, incremental --out-current, incremental --out, loop --out, and promote --out can write the explicitly named artifact.
semantics-core can update the selected Semantics Store because it invokes the core indexer; it still does not write a language profile or ranking.
validate, compare, delta, score, capabilities, coverage-gap, objective-matrix, score-debug-plan, registry, seed, connector, and --mcp-schema are review/read surfaces unless an explicitly documented output option is supplied.
A written review artifact is not a promoted profile, and a promoted-review candidate is not a Monogram ranking change.

Non-Negotiable Boundaries

Benchmark answer files, answer paths, answer symbols, ground_truth.md, and spoiled Monobench output are forbidden profile evidence.
Compiler-only signals are semantic-sitter candidates, not language-wide priors. Promotion requires repetition across at least two language-self corpus kinds.
structure_stats and syntax_shape_stats are lexical proxies. parsed_ast_stats is real Tree-sitter source-AST evidence. Neither is a compiler/type-system oracle.
semantic_flow_stats records local ontology transitions and transfer patterns; it is not call-graph or dataflow proof.
compiler-facts evidence comes through Monogram Semantics and remains an explicit review lane, never a silently persisted ranking signal.

Release Platforms

The 0.1.3 CDN metadata publishes five artifacts. Intel macOS is not present.

Platform key	Binary
`darwin-aarch64`	`monogram-language`
`linux-aarch64`	`monogram-language`
`linux-x86_64`	`monogram-language`
`windows-aarch64`	`monogram-language.exe`
`windows-x86_64`	`monogram-language.exe`

The installed macOS ARM payload is Developer ID signed. Keep that installed identity separate from repository HEAD and from review artifacts generated by the tool.

Relationships

Tool	Ownership
Monogram	Product search, graph, ranking, and score-debug surface
Monogram Semantics	Structural FunctionIR/config-unit store, effects, graphs, bands, compiler-review facts
Monogram Logos	Foundry review over Semantics Store evidence
Monobench	Canary/evaluation telemetry, never profile training data

Get Started

Articles

So Anthropic

User Guide

Philosophy

Applied Philosophy

Architecture

Gestalt Color

Transparency

Terminal Core

Features

VariableMD

Research

CodexMono

OpenCLIs

vpncli

MonoSurf

AI-CLIs

Quick Start

Complete Command Reference

Evidence Flow

Persistence and Review Boundaries

Non-Negotiable Boundaries

Release Platforms

Relationships

​Quick Start

​Complete Command Reference

​Evidence Flow

​Persistence and Review Boundaries

​Non-Negotiable Boundaries

​Release Platforms

​Relationships

Quick Start

Complete Command Reference

Evidence Flow

Persistence and Review Boundaries

Non-Negotiable Boundaries

Release Platforms

Relationships