mneme check enforcement flow, the classes of governance violations the rule engine can operationalize today, supported languages, and the deterministic benchmark methodology.Docs.
Reference for the CLI, the governance violations Mneme catches, supported languages, and the benchmark methodology.
Quick start
Zero to a governed check in three commands. Python 3.11+.
$ pipx install "mneme-hq>=0.5.1" $ mneme init $ mneme check --memory .mneme/project_memory.json --input prompt.txt --query "storage backend"
mneme init creates .mneme/project_memory.json with an empty valid decision corpus (it refuses to overwrite an existing corpus unless --force is supplied). Exit code 0 = pass, 1 = warn, 2 = fail. To enforce on Claude Code Edit and Write calls, install the runtime as above and then load the plugin from a Mneme checkout — runtime and plugin are separate artifacts, and the plugin is not yet in the Claude marketplace: claude --plugin-dir ./integrations/claude-code-plugin. To generate a Cursor rules file: mneme cursor generate --memory .mneme/project_memory.json --query "storage backend".
Reference surfaces
Mneme HQ's documentation is intentionally narrow. We document the surfaces that engineers actually need to operate the governance layer: the CLI you run, the violations Mneme catches, the languages Mneme governs today, and the methodology behind the benchmark. Everything else lives in the source repository.
How Mneme turns architectural decisions into a structured corpus with deterministic enforcement across four workflow boundaries: guidance, prevent, catch, verify — and why retrieval is not enforcement.
How enforcement works →Commands, flags, and exit codes for mneme: init, adr import, list_decisions, add_decision, test_query, check, cursor generate, and benchmark. GitHub Actions and pre-commit CI patterns.
Twelve worked examples across architecture, workflow, security, dependency, and platform governance. Each shows the rule, the AI's offending output, and the structured flag Mneme emits.
Governance violations →Canonical coverage matrix. Tier 1 (Python, TypeScript, JavaScript) and Tier 2 (Go, Java, C#, Rust). Capabilities, limitations, and roadmap per language.
Supported languages →How the v1.1 governance benchmark is measured: layered retrieval and enforcement scoring, structured-output verification, pre-registered thresholds, anti-gaming protocol.
Benchmark methodology →How to read these docs
The CLI reference is the operational entrypoint: every governance action — listing decisions, adding them, running a check, generating Cursor rules, executing the benchmark — flows through one of the documented commands. Pair it with the governance violations doc to understand what the check command is actually surfacing in practice.
The supported languages page is the canonical coverage matrix — Tier 1 (Python, TypeScript, JavaScript) gets native depth; Tier 2 (Go, Java, C#, Rust) gets repository-level governance through architectural and dependency rules. Mneme is language-agnostic by design; that page documents where the depth lives today and what is on the roadmap.
The benchmark methodology page is the methodology specification, not a results dashboard. It describes layered retrieval and enforcement scoring, structured-output verification, pre-registered thresholds, and the anti-gaming protocol — methodology before metrics. Cross-link: see the Insights hub for essays on the underlying architectural-governance category, and the source repository for everything the docs do not cover.