Introduction

Mneme is a proactive engineering intelligence system. It ingests signals from your team's tools — GitHub, Slack, and Linear — builds a memory layer via LLM processing, and enriches pull request analysis with learned context.

The Problem

Every AI coding tool starts from zero. When a new PR is opened, the reviewer (human or AI) has no memory of:

Why a particular architecture decision was made six months ago
That a similar bug was fixed last sprint in a different service
Team conventions around error handling, naming, or testing patterns
Ongoing discussions in Slack about a planned refactor

Mneme solves this by building and maintaining a persistent memory of your engineering team's decisions, patterns, and context.

How It Works

Webhooks (GitHub / Slack / Linear)
    ├── PR events → pr-analysis queue → Analyzer workers → GitHub comment + commit status
    └── Other events → raw_events table → signal-processing queue → Signal workers
                                                                        ├── Store normalized signals + entities
                                                                        ├── Enrich open PR analyses
                                                                        └── memory-processing queue → Memory workers
                                                                              └── Maintenance jobs (decay, consolidation, evolution)

Signal Sources

Mneme ingests signals from three sources:

GitHub — push events, issues, issue comments, PR reviews, review comments
Slack — channel messages from configured channels
Linear — issue creation/updates, comments, label changes

Memory Types

Signals are processed by an LLM and classified into six memory types:

Type	Description	Half-life
Decision	Architecture or design choices	30 days
Pattern	Recurring implementation approaches	20 days
Convention	Team standards and naming rules	60 days
Issue	Bugs, incidents, known problems	7 days
Preference	Individual or team preferences	14 days
Fact	Stable truths about the codebase	90 days

PR Enrichment

When a pull request is opened, Mneme retrieves relevant memories and includes them in the PR analysis. This gives reviewers (and AI tools) the context they need to provide better feedback.

Next Steps

Getting Started — set up Mneme in under 5 minutes
Installation — deploy with Docker or Fly.io
Configuration — customize Mneme via the dashboard settings
Connectors — connect GitHub, Slack, and Linear