felipesauer

Mnema — AI skill for Claude Code

AI community

A signed, append-only record of the decisions behind AI-agent work — the decision, the reasoning, and who wrote it down, in the repository where the work happens.

How to install Mnema

This entry records only its repository, not the path inside it, so there is no exact command to give. Open felipesauer/mnema and copy the folder into ~/.claude/skills/, or the file into ~/.claude/agents/.

What Mnema does

A signed, append-only record of the decisions behind AI-agent work — the decision, the reasoning, and who wrote it down, in the repository where the work happens.

Alternatives in AI

Om Humanize — Voice-calibrated editing — makes Claude-drafted text sound like you wrote it, not like AI wrote it 4.6k ★
Metaharness — 🛠️ The meta-harness for AI agents — scaffold your own focused, branded agent harness with its own npx CLI, MC 616 ★
Upstream Watch — Review pydantic-ai and pydantic-ai-harness releases published since the last watch entry, and record a verdict 361 ★

README

mnema

A signed, append-only record of the decisions behind AI-agent work — the decision, the reasoning, and who wrote it down, in the repository where the work happens.

Tamper-evident, not tamper-proof: what is still in the record has not changed since it was signed, and a stranger can check that without your keys and without installing this.

In one picture

Your project's memory for coding agents: it lives in the repository, it reaches the agent before it writes, and anyone can check it. The reaching is the plugin's: in Claude Code a session is handed the record as it opens and the rules for a file at each edit, and in VS Code and in Cursor's command-line agent it is handed the opening — see [Install](#install).

flowchart LR
    agent["Agent session
Claude Code · VS Code · Cursor CLI"]
    record[(".mnema/
signed · append-only
committed with the code")]
    team["Your team
every clone"]
    stranger["Anyone
no key · no network"]

    agent -- "records decisions, notes, tasks" --> record
    record -- "opens every session with what is in force" --> agent
    record -- "at each edit, in Claude Code: the rules for that file" --> agent
    record -- "git push / clone" --> team
    record -- "mnema verify" --> stranger

An agent decides things all day and leaves almost none of it behind. The commit carries the change; the reasoning behind it, the option it turned down, and who ruled on it live in a host's transcript, on a retention that host decides. mnema puts that half in the repository itself — as typed facts, signed when they are written and hash-chained, so an edit made afterwards cannot be made quietly.

This repository holds the whole product. **[`packages/code`](packages/code/) is the one you install**: the `mnema` command line and its MCP server, two surfaces over one record. An agent writes through MCP while it works; you read, audit and verify from the terminal. Everything else here is what those two surfaces stand