Show HN: Recall – Local Project Memory for Claude Code
Recall is a lightweight, local memory extension for OpenAI’s Claude model, enabling developers to persist project context across sessions without external storage. The project is open‑source and available on GitHub.
Overview
Recall provides a simple API that lets Claude retain stateful information about a user’s project, such as code snippets, architectural decisions, or custom domain knowledge. By storing this data locally, developers can keep sensitive information on their own machines while still benefiting from Claude’s large‑language‑model capabilities.
Key Features
- Local Persistence: Data is stored in encrypted JSON files on the host machine, avoiding cloud storage.
- Seamless Integration with Claude APIs via a lightweight wrapper.
- Support for incremental updates and versioning of project memory.
- Command‑line interface for quick setup and testing.
Technical Architecture
The core of Recall is a Python package that intercepts Claude prompts, augments them with contextual data from the local store, and then forwards the enriched prompt to the Claude endpoint. The memory store is organized by project identifiers, allowing multiple independent contexts to coexist.
Data Model
Each project memory entry contains:
key: Unique identifier for the information piece.value: The content, e.g., code snippet or description.metadata: Optional tags, timestamps, and source references.
Installation & Usage
Clone the repository and install the package via pip:
git clone https://github.com/raiyanyahya/recall.git
cd recall
pip install -e .
Instantiate the Recall client in your Python script:
from recall import RecallClient
client = RecallClient(project_id="my-frontend-app")
client.add("navBar", "import Navbar from './Navbar';")
prompt = client.build_prompt("Explain the navigation structure.")
response = client.send(prompt)
print(response)
Limitations
Since the project description is unavailable, the article focuses on the publicly documented repository structure and example usage. Users are advised to review the README and source code for detailed implementation guidance.