iworkforces

ReflectLog — AI skill for Claude Code

AI community

An Agentic Memory Layer For AI Agents.

How to install ReflectLog

This entry records only its repository, not the path inside it, so there is no exact command to give. Open iworkforces/ReflectLog and copy the folder into ~/.claude/skills/, or the file into ~/.claude/agents/.

What ReflectLog does

An Agentic Memory Layer For AI Agents. Including Grok, Claude Code, Codex, OpenCode, OpenClaw and Hermes.

Alternatives in AI

EchoBird — One-click install + model switch:Claude Code,Codex CLI (OpenAI), Grok Build (xAI), DeepSeek Harness, Kimi Code 3.1k ★
Token Monitor — Local-first desktop widget for tracking token usage, costs, and limits across 32+ AI coding tools—including Cl 1.8k ★
Iai Personal Memory Engine — A cyber brain for your AI 866 ★

README

ReflectLog

An Agentic Memory Layer For Coding Agents

[![Python](https://img.shields.io/badge/python-%E2%89%A53.14.4-blue)](https://www.python.org/) [![License](https://img.shields.io/badge/license-MIT-green)](LICENSE)

ReflectLog is an [MCP (Model Context Protocol)](https://modelcontextprotocol.io) server that provides persistent, Workspace-based memory storage for Claude Code and other AI agents. It combines semantic vector search with full-text search for intelligent memory retrieval.

Features

Hybrid Search: Combines semantic similarity (USearch) + exact phrase matching (Tantivy)
RRF Fusion: Reciprocal Rank Fusion for optimal result ranking
Pluggable Reranking: Local cross-encoder relevance scoring (or none)
Temporal-Aware Scoring: Recency decay for handling contradictory memories
Smart Memory Replacement: LLM-based detection of memory updates
HTTP Transports: http (default) and streamable-http, with bearer-token auth
Eager engine warmup by default, with optional lazy reranker load

Quick Start

Installation

# Clone the repository
git clone https://github.com/iWorkforces/ReflectLog.git
cd ReflectLog

# Install dependencies using uv
uv sync

Configuration

Create a `.env` file with your API key:

OPENROUTER_API_KEY=sk-or-your-key-here

Running the Server

# Start with the default HTTP transport (requires MCP_AUTH_TOKEN; bind-all needs ALLOW_PUBLIC_BIND=true)
MCP_AUTH_TOKEN=change-me uv run reflectlog --port 9103

# Or use the launcher
MCP_AUTH_TOKEN=change-me ./start-reflectlog-mcp-server.sh

The server accepts only the `http` (default) and `streamable-http` transports. `stdio` and `sse` are not supported.

Connect HTTP MCP clients

Choose a secret token and use the same value for the server's `MCP_AUTH_TOKEN` and each client's `REFLECTLOG_MCP_TOKEN`. For example, replace `change-me` in the command above and export the matching value in the environment t