AI agents lose context every session. repomemory fixes that — one command creates a persistent, searchable knowledge base that any AI tool can read, search, and write to.
Persistent memory for AI coding agents. Hybrid search, auto-session capture, context routing.
2026-11-09
星數
—
工具
—
可用率
—
npx repomemory go
The Problem
Every time you open a project with Claude Code, Cursor, Copilot, or any AI coding agent:
It re-discovers your architecture from scratch
It proposes changes that were already debated and rejected
It re-introduces bugs that were already fixed
Your CLAUDE.md / .cursorrules helps, but it's static and gets stale.
The Solution
.context/
├── index.md ← Quick orientation (loaded every session)
├── facts/ ← Architecture, database, deployment
├── decisions/ ← "We chose Drizzle over Prisma because..."
├── regressions/ ← "This broke before. Here's what happened."
├── preferences/ ← Your coding style — follows you across all repos
├── sessions/ ← Auto-captured AI session summaries
└── changelog/ ← Monthly git history syncs
Facts tell agents how things work. Decisions prevent re-debating. Regressions prevent re-breaking. Preferences teach agents how you code.
Quick Start
With an API key (Claude Code, terminal workflows)
npx repomemory go
One command: sets up global profile, creates .context/, configures Claude Code + Cursor, runs AI analysis, prints CLAUDE.md instructions.
With Cursor (no API key needed)
npx repomemory setup cursor
This installs everything Cursor needs:
MCP server in ~/.cursor/mcp.json (auto-starts repomemory in every project)
Rules in .cursor/rules/repomemory.mdc (teaches Cursor's AI to use context)
6 commands in .cursor/commands/ (run with / in Cursor chat)
Then in Cursor chat, type:
/repomemory-analyze
Cursor's own AI scans your repo and populates .context/ via the MCP tools. No external API key required — your Cursor subscription handles it.
Available Cursor commands:
Command
What it does
/repomemory-analyze
Full repo analysis — populates facts, decisions, index
/repomemory-orient
Quick orientation at start of session
/repomemory-search
Search the knowledge base
/repomemory-record
Record a fact, decision, or regression
/repomemory-session
Save a session summary
/repomemory-status
Show context coverage
Guided wizard
npx repomemory wizard
Walks through provider selection, tool integration, and first analysis. If no API keys are detected, offers a "None — I use Cursor" option that skips external analysis entirely.
Non-interactive (CI-safe)
npx repomemory go --yes --provider anthropic --embedding-provider gemini --max-files 80
No prompts when --yes / --defaults / --no-prompt is used.
MCP Server — Agents With Real Memory
The real power is the MCP server. When configured via repomemory setup claude, it auto-starts with Claude Code and gives agents 7 tools:
Tool
What It Does
context_search
Hybrid keyword + semantic search across repo + global context
Write entries with smart scope routing (preferences → global)
context_read
Read full content, repo-first with global fallback
context_list
Browse entries with [repo]/[global] provenance tags
context_risk
Assess modification risk — hotspots, hidden coupling, bus factor
context_delete
Remove stale knowledge
Agent: "Let me orient myself in this project..."
→ context_auto_orient()
→ Returns: project overview, preferences, recent sessions, recent changes
Agent: "Let me search for context about the auth flow..."
→ context_search("authentication flow")
→ Auto-routes to facts/ category, returns compact results
Agent: "I found a race condition. Let me record this."
→ context_write(category="regressions", filename="token-refresh-race", content="...")
→ Persisted. Detects if it supersedes an existing entry.
Sessions are auto-captured on shutdown. Zero config — repomemory setup claude handles everything.
Supported Tools
Tool
Integration
API Key Required?
Claude Code
MCP server (auto-starts) + post-commit hook
Yes (for analysis)
Cursor
MCP server + rules + 6 slash commands
No — uses Cursor's built-in AI
GitHub Copilot
copilot-instructions.md
Yes (for analysis)
Windsurf
.windsurfrules
Yes (for analysis)
Cline
.clinerules
Yes (for analysis)
Aider
.aider.conf.yml
Yes (for analysis)
Continue
.continue/rules/
Yes (for analysis)
Cursor users: You don't need any API key. Run npx repomemory setup cursor, then use /repomemory-analyze in Cursor chat. Cursor's AI does the analysis using the MCP tools — whatever model Cursor is using (it can even switch models mid-task).
Supported Providers
Provider
Models
Env Variable
anthropic
claude-sonnet-4-6, claude-opus-4-6
ANTHROPIC_API_KEY
openai
gpt-4o, o3-mini
OPENAI_API_KEY
gemini
gemini-2.0-flash, gemini-2.5-pro
GEMINI_API_KEY / GOOGLE_API_KEY
grok
grok-3, grok-3-mini
GROK_API_KEY / XAI_API_KEY
Embeddings (optional, for semantic search): Gemini text-embedding-004 (free, default) or OpenAI text-embedding-3-small. Auto-detected from available API keys.
All Commands
repomemory go # One-command setup (add --yes for non-interactive)
repomemory wizard # Interactive guided setup
repomemory analyze # AI-powered repo analysis
repomemory analyze --merge # Update without overwriting edits
repomemory analyze --dry-run # Preview without API call
repomemory search <query> # Search knowledge base from terminal
repomemory status # Coverage and freshness report
repomemory doctor # Diagnostics and health check
repomemory dashboard # Local web UI (localhost:3333)
repomemory risk # Hotspots, coupling, ownership analysis
repomemory risk -f src/auth.ts # Targeted risk assessment for specific files
repomemory sync # Sync git history to changelog
repomemory setup <tool> # Configure Claude/Cursor/Copilot/etc
repomemory hook install # Auto-sync changelog on commits
repomemory global list # Manage global developer context
repomemory global export # Export global context as JSON
Configuration
.repomemory.json in your repo root (all fields optional):
