Context, Memory & RAG tools for Claude Code

Agent memory system organized around three explicit operations—retain, recall, and reflect—with semantic, keyword, graph, and temporal retrieval plus an MCP server. The June 2026 release and production usage make it a concrete reference for turning cross-session persistence from passive storage into an active learning layer inside the harness

Persistent memory server for coding agents that retrieves only the top-K relevant observations via hybrid BM25 + vector + graph search instead of loading a whole memory file into context (~1,900 vs 22K+ tokens at 240 observations, 92% fewer). Reports 95.2% R@5 on LongMemEval-S with a local embedding model, plus a reproducible eval harness to check that claim. One server serves 20+ agents (Claude C

Agent Package Manager

Mounts S3, Slack, Gmail, GitHub, and Redis side-by-side as a single virtual filesystem so agents interact with every backend through familiar bash commands instead of learning N distinct APIs. The key harness insight: LLMs are already fluent in grep, cat, and cp — leveraging that vocabulary eliminates tool-schema bloat and makes cross-service pipelines compose as naturally as local shell scripts

Agents that self-improve with experience

Portable project memory across Claude Code, Codex and OpenCode, plus token accounting measured from harness transcripts. Local file I/O, no API calls, no telemetry.

Turns agent-learned project knowledge into a repo-local, symbol-grounded wiki with drift detection and task-aware routing. The critical gap it fills: most agent memory stores facts, but mex keeps those facts connected to the exact code symbols they describe and flags when code changes invalidate them — making memory freshness a harness-level concern rather than a manual cleanup task

by activeloopai - Hivemind turns your traces into reusable skills across agents

Self-improving executable context layer for data and analytics agents: it ingests warehouses, BI tools, and wikis to build a semantic layer with approved metrics, joinable columns, and resolved fan/chasm traps, then serves the result to Claude Code, Codex, and Cursor through MCP. Fills the gap where general-purpose agents invent metric logic on every question — it turns warehouse querying from a p

Turns raw Claude Code sessions into a self-evolving knowledge base: hooks capture every interaction, the Agent SDK extracts decisions and lessons, and an LLM compiler distills them into structured, cross-referenced articles that improve retrieval quality over time. The most concrete open-source implementation of trace-driven memory evolution for coding agents

Repository-level operating harness that turns any software repo into an agent-ready workspace: structured AGENTS.md, HARNESS.md, and FEATURE_INTAKE.md give agents the missing project context — where to start, what the product contract says, how risky the change is, and which decisions future agents should inherit. The most concrete open-source implementation of "coding agents need better repositor

Turns corrections into Preferences, Project-specific skills, and Shared skills for Claude Code, Codex, and OpenCode.

AI 应用开发、AI 编程实战与面试指南，涵盖 LLM、Agent、RAG、MCP、Claude Code、Codex 等核心技术与工程实践。

Frozen — cross-agent persistent memory for coding assistants. Still works; the compressed-memory core now ships inside JuliusBrussee/caveman.

A persistent, file-based memory layer for coding agents — give Claude Code, Codex & others a project brain (durable decisions, requirements, constraints) via a zero-dependency CLI.

by David Corvoysier - Durable cross-agent memory for Claude Code, Codex, pi, and Hermes. Your memory is a Hugging Face dataset you own: recall it from any machine or share it with your team. Single binary, local index, pinned local embedding models

Notes for you, Memory for your agents. / 内置 Deepseek harness Agent / 适用 办公 & 写作 & Coding

Open-source, Git-native memory for AI coding agents — deterministic local recall for Claude Code, Codex, Cursor, Gemini CLI, and compatible tools.

memX: self-learning, self-maintaining memory plugin for AI agents; native support for claude code, codex, and openclaw

AI builder operating system. Turn Claude Code from a per-session tool into a workspace that remembers, maintains itself, and never goes stale. BASE = Builder's Automated State Engine.

Markdown-first shared memory vault for Claude Code and Codex with SQLite, Zvec, Git, closeout, and audit

by plur-ai - Open, local-first engram memory for Claude Code. Engrams are plain-text YAML you can read, correct, and delete. The same memory is shared across MCP tools and multiple providers

Letta's July 2026 library that normalizes native session transcripts from 15+ harnesses (Claude Code, Codex, Cursor, OpenCode, Pi, Gemini CLI, OpenHands, and more) into one validated, model-ready record format. Fills a real infrastructure gap — every runtime logs the same concepts (messages, reasoning, tool calls, results) in incompatible formats — and is explicitly designed for *agent* consumptio

Memory component for agents that keeps each memory with its source evidence in one local SQLite file and retires memories recorded as depending on a source when that source is corrected; TypeScript SDK, stdio MCP server, CLI and HTTP API, with exact replay and bisect in the 3.0 pre-release. MIT

by Egor Fedorov - Claude Code plugin that tracks token usage, identifies wasted context, and saves money on unnecessary API costs, by tracking which information is actually being reused later. Visuals include heatmaps, ROI reports, budget alerts, efficiency scores, git-aware suggestions - all local, zero config. The design is still somewhat exploratory, but it shows promise

Engineering discipline layer for Claude Code -- 67 slash commands, 20 coding rules, 27 skills, 9 agents, quality gates between every step. 8 stack profiles (Go, Python, Java, Kotlin, React, etc.), agent memory across sessions. Install: npx @c0x12c/spartan-ai-toolkit@latest --local

Local dual-layer memory pattern for AI agents: a compact, human-readable markdown index paired with semantic retrieval from a local vector store, queried before each message. For cross-project recall where flat memory files or vector-only RAG fall short

by Wolfe-Jam - Authors and versions CLAUDE.md and AGENTS.md from a repository's actual detected stack, so Claude Code's project context stays in sync with the code instead of drifting. One typed source file (project.faf, an IANA-registered media type) is the single origin for every assistant's context file; a deterministic score reports how complete that context is. The file lives in the repo and

Drop-in 12-rule CLAUDE.md + AGENTS.md baseline that closes common agent-orchestration failures (token spirals, silent partial failures, two-pattern pollution). Includes PRD generator prompt, browser-skill-graduation workflow, and 5 example skills. ~700 tokens total, MIT

by lukaszraczylo - Memory management and retrieval for Claude Code

by jungjaehoon-lifegamez - Always-on companion for Claude that remembers your decisions and their evolution. Local-first memory using SQLite + transformers.js embeddings

by serpro69 - 🦫 Privacy-first virtualization layer for LLM context with MCP protocol for tool access

Open-source linter for AI context files (CLAUDE.md, .cursorrules, copilot-instructions.md) that catches stale paths, wrong commands, and token waste by validating against the real codebase

by sara-star-quant - Per-repo memory, outcome telemetry, and a calibrated-confidence gate for Claude Code, with MCP and AGENTS.md projections so other AI coding tools can read its context. Notes survive sessions; success claims need test evidence; your reverts are remembered. Local-only, stdlib runtime

One source of truth for AI coding rules across every IDE. Author rules, agents, and skills once in plain markdown, and the engine routes them into each tool's native format (Claude Code, Cursor, Copilot, Codex, Pi, OpenCode, AGENTS.md) with no duplication or drift. Zero dependencies, bash + awk, MIT

CLI for managing Claude Code project configs across custom providers (Databricks, Bedrock, OpenRouter, LiteLLM, Cloudflare). Overlay merge/remove, MCP web search setup, multi-provider switching, team config export/import

CLI and MCP server that generates AGENTS.md/CLAUDE.md from static analysis of codebase structure, git history, and convention detection. Surfaces hub files, co-change coupling, and blast radius. No LLM required

Single-file Python linter that scores any CLAUDE.md / AGENTS.md against a 12-rule baseline. Flags leaked secrets (GitHub PATs, AWS keys, PayPal links), the 200-line compliance cliff, and missing project-specifics sections. Zero dependencies, JSON output for CI, MIT

Cross-agent registry and CLI for discovering and installing 220+ AI assets including skills, prompts, MCP configs, and workflows