A new tool called Recall appeared on Hacker News offering fully local project memory for Claude Code. The post received 59 points and 49 comments.
Recall stores conversation context and project details on the user's machine instead of sending them to external servers.
What It Is and How It Works
Recall runs entirely on the developer's computer. It indexes code files, previous prompts, and Claude responses within a local database. When a new session starts, the tool retrieves relevant snippets and injects them into the prompt sent to Claude Code.
No API keys or cloud accounts are required for the memory layer. The system uses standard local file scanning and embedding generation to keep everything offline.
Community Metrics from the Launch
The Hacker News thread shows 59 points and 49 comments within the first day. Early discussion focused on privacy benefits and questions about embedding model size.
No official benchmark numbers were published in the announcement. Users in the thread reported typical context retrieval times under two seconds on mid-range laptops.
How to Try It
Clone the repository and follow the setup instructions in the README. The tool requires Python 3.10+ and a local embedding model such as all-MiniLM-L6-v2.
After installation, point Recall at a project folder and start a Claude Code session. The memory layer activates automatically on subsequent prompts.
Pros and Cons
- Keeps all project context on disk with no data leaving the machine
- Works offline after initial model download
Integrates directly with existing Claude Code workflows
Requires local compute for embeddings
Limited to file-based projects; no support for remote repositories yet
No published accuracy benchmarks against cloud memory solutions
Alternatives and Comparisons
Developers currently use several approaches for Claude Code memory.
| Feature | Recall (local) | Claude Projects (cloud) | Custom RAG scripts |
|---|---|---|---|
| Data location | Local disk | Anthropic servers | User controlled |
| Setup complexity | Medium | Low | High |
| Offline capability | Yes | No | Yes |
| Cost | Free | Subscription | Free |
Recall sits between fully manual RAG setups and Anthropic's hosted project feature.
Who Should Use This
Teams handling sensitive codebases or working under strict data residency rules gain the most. Individual developers who already run local LLMs will find the integration straightforward.
Users who rely on Claude's built-in project memory and do not mind cloud storage can skip it.
Bottom Line / Verdict
Recall delivers the first practical, fully local memory layer for Claude Code without requiring custom infrastructure.
The tool lowers the barrier for developers who need persistent context while keeping data on their own hardware. Its adoption will depend on how quickly the community adds features such as multi-project support and better retrieval accuracy.
Top comments (0)