Lore.AI
LLM proxy for coding agent context and memory management
Last verified:
What is Lore.AI?
Lore.AI — ore AI is a persistent memory and shared context system for AI agents that solves the problem of context loss across AI coding sessions. Instead of losing information when AI tools compact conversations or when developers start fresh sessions, Lore intercepts every message between your AI client and the upstream API, distilling conversations into timestamped observation logs that preserve decisions, file paths, patterns, and the
Lore.AI pricing
Pricing model: Freemium
Lore is free and local today. The tool costs nothing to use—you bring your own API keys for OpenAI and Anthropic. Background distillation and curation run at 50% off via batch APIs on cheaper models. There are no paid plans currently; the only paid cost is your own API usage with OpenAI or Anthropic. Folk Lore (team memory sharing) is in early access with rolling availability but pricing details are not yet announced.
Lore.AI pros
- Persistent shared context across AI sessions lasting days and hundreds of turns
- Replaces lossy compaction with timestamped observation logs preserving all details
- Automatic extraction of technical learnings and decision rationales without manual work
- On-device vector search with Nomic Embed v1.5—zero API cost and zero latency
- 2.6x total recall compared to compaction at 2.3M tokens
- Preserves the
- behind every choice preventing AI from refactoring workarounds
- Five feedback loops for behavioral pattern detection and adaptive calibration
- Import existing conversations from multiple AI tools so you don't start blank
- Works with any Anthropic or OpenAI-compatible tool without client changes
- Cost-aware caching with 50% off via batch APIs on cheaper models
- Project-specific and global memory with confidence-ranked priority injection
- Predictive cache warming avoids expensive cache rebuilds that compaction triggers
- Fair source local engine keeping all data on your machine
- No context files to write or instructions to maintain
- Memory travels with you across providers, tools, and machines without vendor lock-in
Lore.AI cons
- Currently free and local only—Folk Lore team sharing is still in early access
- Requires changing base URL in AI client configuration
- Only supports Anthropic and OpenAI-compatible protocols
- On-device search requires local hardware for vector embedding
- No cloud-based sharing for teams until Folk Lore launches
- Might not work with AI tools using non-standard APIs
- Local installation needed rather than pure cloud service
- Early access rolling out for team features means limited collaboration now
Frequently asked questions about Lore.AI
What problem does Lore AI solve?
Lore solves context loss in AI coding sessions. When AI tools compact conversations or developers start fresh sessions, the AI forgets decisions, file paths, and the reasoning behind choices. This leads to worse answers, undone decisions, and hours spent re-explaining your project. Lore captures every message and distills it into timestamped observation logs, preserving persistent shared context across sessions lasting days and hundreds of turns.
How does Lore work with my existing AI tools?
Lore sits between your AI client and the upstream API, intercepting every message. You only need to change your base URL to Lore's endpoint—no client changes needed. It works with Claude Code, OpenCode, Pi, Codex, and any tool that speaks Anthropic or OpenAI-compatible protocols including Aider, Cline, and Continue.
How much better is Lore than compaction?
In a real 5-day coding session with 2.3 million tokens, Lore scores 4.0/5 on recall while compaction scores 2.4/5. Lore delivers 2.6x total recall over compaction with 13 perfect scores versus 5. Developers save 68 minutes per day on re-explaining because Lore remembers for you.
What is the on-device vector search feature?
Lore runs Nomic Embed v1.5 locally for vector search with zero API cost and zero latency. It uses a hybrid architecture fusing vector similarity with BM25 keyword search for best-of-both-worlds recall. LLM-powered query expansion finds what you mean, not just what you typed, making details from every session searchable even hundreds of turns later.
Can I import my existing AI conversations?
Yes. Lore imports conversations from Claude Code, Codex, Aider, Cline, Continue, OpenCode, and Pi. Running lore import auto-detects these tools and extracts knowledge from your existing sessions so your AI starts with context from day one instead of a blank slate.
Is Lore cloud-based or local?
Lore runs locally with a fair source engine, keeping all your data on your machine. There's no vendor lock-in or walled gardens. You can switch providers, tools, or machines and your memory travels with you. The only cloud component is your own API usage with OpenAI or Anthropic.
What is Folk Lore and when is it available?
Folk Lore brings your team's memory together—shared, searchable, and always current across team members. It's the collaboration feature where every conversation makes every future session smarter across any provider, tool, or team member. Folk Lore is currently in early access with rolling availability, but pricing details are not yet announced.