totem, from Mmnto Ai, is an open-source, file-based toolkit that enforces architectural guardrails for AI coding agents. It provides deterministic code checking and a local, queryable knowledge index so agents reference project rules during development. The CLI exposes commands such as lint, doctor, and lesson-compile; the tool runs without Node.js and integrates into CI/CD pipelines. Software engineers, DevOps professionals, and teams using AI assistants gain repeatable policy enforcement inside automated workflows.

What tasks can you actually use it for?

Use the tool to codify project rules, convert past errors into reusable lessons, and verify AI-produced code against those rules. The system targets common development jobs: offline rule validation before merges, automated checks in pipelines, and converting historical mistakes into formal checks that agents can consult. Practical outputs are lint-style reports and a queryable index of lessons that agents reference during code generation.

How reliable are its rule checks?

The linting logic applies deterministic decision rules so the same inputs produce the same checks on repeated runs. That determinism makes enforcement predictable in automated runs, but it does not guarantee the generated code is semantically correct in every case. The usefulness of results therefore depends on the quality and coverage of the authored lessons compiled into the knowledge index.

What inputs and environments does it require?

The system uses a local, file-based knowledge repository and a command-line interface that runs on Linux, macOS, and Windows. It supports offline operation for rule evaluation and requires project files and authored lesson documents as its primary inputs. Key CLI commands include:

• totem lint for offline checks
• totem doctor for troubleshooting rule conflicts
• totem lesson compile to turn experiences into queryable rules

Is it practical for team workflows and automation?

Totem integrates into CI/CD pipelines to enforce standards across team contributions and fits repository-based workflows because its knowledge store is file-first. The local-first model keeps architectural rules in source control, simplifying sharing among engineers. Adoption requires time to author the rule corpus; teams that treat lessons as part of their development artifacts can enforce reproducible agent behaviour across branches and automated checks.

A pragmatic choice for teams ready to maintain rule corpora

totem is a pragmatic option for engineering teams seeking reproducible enforcement of agent behaviour, backed by its open-source distribution and recognition among early MCP adopters. The main trade-off is ongoing maintenance: the tool’s value rises with the effort invested in curating lesson and rule sets. Teams prepared to treat governance as code will find it a durable aid to consistent AI-assisted development.