bootspring memory
Extract and search project learnings from git history, plus cross-session memory recall.
Overview
The memory command has two layers:
- Git-based memory -- Analyzes your git history to extract learnings, patterns, and decisions
- Cross-session recall -- Searches
.bootspring/memory/sessions/for past decisions and context saved across AI assistant sessions
Usage
Loading code block...
Commands
| Command | Description |
|---|---|
summary | Show categorized learning summary (default) |
search <query> | Search learnings by keyword |
recall <query> | Search cross-session memory for past decisions |
save <note> | Save a decision or note to cross-session memory |
files | Show most-changed files |
stats | Show repository statistics |
categories | List memory categories |
export | Export learnings as JSON |
Options
| Option | Scope | Description |
|---|---|---|
--limit <n> | summary, search, files, export | Number of commits to analyze (default: 30) |
--since <date> | summary, export | How far back to look (default: "3 months ago") |
--json | recall | Output as JSON |
--tag <tag> | save | Tag for categorization |
Categories
Commits are auto-categorized based on their message:
| Category | Label | Matches |
|---|---|---|
| bugfix | Bug Fixes | fix:, bugfix:, messages containing "fix" |
| feature | Features | feat:, feature:, messages containing "add" |
| refactor | Refactoring | refactor:, messages containing "refactor" |
| security | Security | security:, messages containing "security", "auth" |
| performance | Performance | perf:, messages containing "performance", "optimize" |
| docs | Documentation | docs:, messages containing "document" |
| test | Testing | test:, messages containing "test" |
| build | Build | build:, ci:, messages containing "build" |
| style | Style | style:, messages containing "format", "lint" |
| chore | Chores | chore:, messages containing "chore" (default) |
Examples
Git-Based Memory
Loading code block...
Cross-Session Recall
Loading code block...
Hook-Based Memory Agent
Bootspring includes a persistent memory agent that runs via Claude Code hooks. When hooks are installed (bootspring hook install), structured context blocks are injected into every conversation automatically:
- persona -- Your name, role, communication style
- project_context -- Repo name, tech stack, structure (auto-populated)
- user_preferences -- Workflow preferences, commit conventions
- session_notes -- Carry-over notes from previous sessions
See the Hooks Guide and Session Intelligence for details.
Related Commands
bootspring hook- Hook configuration (includes memory init)bootspring session- Session snapshotsbootspring observe- Session observability