How Mneme Works
How Mneme turns your team's activity into persistent memory for AI tools.
Mneme watches your team's tools — GitHub, Slack, Linear — and builds a persistent memory layer that your AI coding tools can query during development.
The problem
Every AI coding tool starts from zero. When you open a file, your AI assistant has no idea that:
- The team decided to use event sourcing for this service last month
- A similar bug was fixed last sprint in a different repo
- There's a Slack thread about a planned refactor of this module
- Your team convention is to use explicit error types, not generic
Error
Mneme fills that gap.
How it works
1. Install the GitHub App
Connect Mneme to your GitHub organization. Optionally add Slack, Linear, or Jira from your dashboard's connectors page.
2. Signals flow in
Every PR, commit, issue, Slack message, or Linear update becomes a signal — a normalized event with extracted entities (people, repos, files, teams).
3. Memories build up
An LLM processes each signal and classifies it into one of six memory types:
| Type | What it captures | Example |
|---|---|---|
| Decision | Architecture or design choices | "Team decided to use event sourcing for orders" |
| Pattern | Recurring implementation approaches | "Error boundaries are placed at route level" |
| Convention | Team standards and naming rules | "All API endpoints use camelCase responses" |
| Issue | Bugs, incidents, known problems | "Race condition in the payment webhook handler" |
| Preference | Individual or team preferences | "Team prefers explicit error types over generic Error" |
| Fact | Stable truths about the codebase | "Auth service uses JWT with RS256 signing" |
Memories aren't static — they fade over time if not reinforced, so the system naturally stays current. When multiple signals corroborate the same memory, its confidence grows stronger.
4. AI tools query via MCP
Connect your IDE to Mneme's MCP server. When you're working on code, your AI assistant can search your team's memories for relevant context — architecture decisions, past bugs, team conventions — without you having to explain it.
What you see in the dashboard
- Signals — a real-time feed of activity from your connected tools, with timeline and filtering
- Memories — the distilled knowledge Mneme has learned, with confidence scores and evidence trails
- Entity hub — click any person, repo, or file to see all related activity
- Settings — manage team access, API keys, and integrations
Next steps
- MCP server — connect your AI coding tools to your team's memories