[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"project-77521":3},{"id":4,"name":5,"fullName":6,"owner":7,"repo":5,"description":8,"homepage":9,"htmlUrl":10,"language":11,"languages":10,"totalLinesOfCode":10,"stars":12,"forks":13,"watchers":14,"openIssues":15,"contributorsCount":16,"subscribersCount":16,"size":16,"stars1d":17,"stars7d":18,"stars30d":19,"stars90d":16,"forks30d":16,"starsTrendScore":20,"compositeScore":21,"rankGlobal":10,"rankLanguage":10,"license":10,"archived":22,"fork":22,"defaultBranch":23,"hasWiki":24,"hasPages":22,"topics":25,"createdAt":10,"pushedAt":10,"updatedAt":32,"readmeContent":33,"aiSummary":34,"trendingCount":16,"starSnapshotCount":16,"syncStatus":15,"lastSyncTime":35,"discoverSource":36},77521,"memX","NeoLi00\u002FmemX","NeoLi00","memX: self-learning, self-maintaining memory plugin for AI agents; native support for claude code, codex, and openclaw","",null,"TypeScript",277,3,7,2,0,34,57,215,102,88.81,false,"main",true,[26,27,28,29,30,31],"agent","agent-memory","embeddings","graph-memory","long-term-memory","openclaw","2026-06-12 04:01:21","\u003Cp align=\"center\">\n  \u003Cimg src=\".\u002Fassets\u002Fmemx-cover-en.svg\" alt=\"memX - self-learning, self-maintaining memory for AI agents\" width=\"920\">\n\u003C\u002Fp>\n\n\u003Cp align=\"center\">\n  \u003Ca href=\".\u002FREADME.md\">English\u003C\u002Fa> · \u003Ca href=\".\u002FREADME-ch.md\">中文\u003C\u002Fa> ·\n  \u003Ca href=\".\u002FARCHITECTURE.md\">Architecture\u003C\u002Fa>\n\u003C\u002Fp>\n\n---\n\nmemX turns completed work into structured, searchable, self-maintained memory, then injects only the evidence an agent needs for the current query.\nIt connects natively to Codex, Claude Code, and OpenClaw, and reaches any MCP-compatible client through the same local memory layer.\n\n## Benchmarks\n\n\u003Ctable align=\"center\">\n  \u003Cthead>\n    \u003Ctr>\n      \u003Cth>Suite\u003C\u002Fth>\n      \u003Cth>Scope\u003C\u002Fth>\n      \u003Cth>R@3 success rate\u003C\u002Fth>\n    \u003C\u002Ftr>\n  \u003C\u002Fthead>\n  \u003Ctbody>\n    \u003Ctr>\n      \u003Ctd>\u003Cstrong>LongMemEval-S\u003C\u002Fstrong>\u003C\u002Ftd>\n      \u003Ctd>Long-context memory retrieval\u003C\u002Ftd>\n      \u003Ctd>\u003Cstrong>94.2%\u003C\u002Fstrong>\u003C\u002Ftd>\n    \u003C\u002Ftr>\n    \u003Ctr>\n      \u003Ctd>\u003Cstrong>Real engineering cases\u003C\u002Fstrong>\u003C\u002Ftd>\n      \u003Ctd>30 cases, each with 20+ turns\u003C\u002Ftd>\n      \u003Ctd>\u003Cstrong>100%\u003C\u002Fstrong>\u003C\u002Ftd>\n    \u003C\u002Ftr>\n  \u003C\u002Ftbody>\n\u003C\u002Ftable>\n\n## Architecture\n\n\u003Cp align=\"center\">\n  \u003Cimg src=\".\u002Fassets\u002Fmemx-overview.svg\" alt=\"memX coarse architecture\" width=\"920\">\n\u003C\u002Fp>\n\n## Agent support\n\n\u003Ctable align=\"center\">\n  \u003Ctr>\n    \u003Ctd align=\"center\" width=\"56\">\u003Cimg src=\".\u002Fassets\u002Fagent-logos\u002Fcodex.png\" alt=\"Codex logo\" width=\"34\">\u003C\u002Ftd>\n    \u003Ctd>\u003Cstrong>Codex\u003C\u002Fstrong>\u003C\u002Ftd>\n    \u003Ctd>\u003Csub>native hooks, MCP hidden by default\u003C\u002Fsub>\u003C\u002Ftd>\n  \u003C\u002Ftr>\n  \u003Ctr>\n    \u003Ctd align=\"center\" width=\"56\">\u003Cimg src=\".\u002Fassets\u002Fagent-logos\u002Fclaude-code.png\" alt=\"Claude Code logo\" width=\"34\">\u003C\u002Ftd>\n    \u003Ctd>\u003Cstrong>Claude Code\u003C\u002Fstrong>\u003C\u002Ftd>\n    \u003Ctd>\u003Csub>native hooks, MCP hidden by default\u003C\u002Fsub>\u003C\u002Ftd>\n  \u003C\u002Ftr>\n  \u003Ctr>\n    \u003Ctd align=\"center\" width=\"56\">\u003Cimg src=\".\u002Fassets\u002Fagent-logos\u002Fopenclaw.png\" alt=\"OpenClaw logo\" width=\"34\">\u003C\u002Ftd>\n    \u003Ctd>\u003Cstrong>OpenClaw\u003C\u002Fstrong>\u003C\u002Ftd>\n    \u003Ctd>\u003Csub>native + hooks\u003C\u002Fsub>\u003C\u002Ftd>\n  \u003C\u002Ftr>\n  \u003Ctr>\n    \u003Ctd align=\"center\" width=\"56\">\u003Cstrong>MCP\u003C\u002Fstrong>\u003C\u002Ftd>\n    \u003Ctd>\u003Cstrong>MCP clients\u003C\u002Fstrong>\u003C\u002Ftd>\n    \u003Ctd>\u003Csub>any MCP-compatible client\u003C\u002Fsub>\u003C\u002Ftd>\n  \u003C\u002Ftr>\n\u003C\u002Ftable>\n\n## Quick start\n\nRequirements: Node.js 22.14+ or Node 24. OpenClaw installs require OpenClaw 2026.3.25+. Python 3 is\nneeded only for the default local embedding runtime.\n\nThe README commands use the GitHub package spec. A fresh run pulls current GitHub code, so installs\ndo not wait for an npm publish. To use the npm release channel later, replace\n`github:NeoLi00\u002FmemX` with `@neoli00\u002Fmemx`.\n\nFill in these values before running a command:\n\n- `--llm-provider`: the provider adapter memX should call. Choose one of `openai-compatible`,\n  `anthropic`, `google`, or `ollama`.\n- `--llm-base-url`: the base URL for that provider. Examples: `https:\u002F\u002Fapi.openai.com\u002Fv1`,\n  `https:\u002F\u002Fapi.anthropic.com\u002Fv1`, `https:\u002F\u002Fgenerativelanguage.googleapis.com\u002Fv1beta`, or\n  `http:\u002F\u002F127.0.0.1:11434` for Ollama.\n- `--llm-model`: the model memX uses for memory compilation, recall planning, and maintenance.\n  Pick a fast, low-cost model with reliable JSON output.\n- `--llm-api-key`: the API key for the provider. Use `--llm-api-key-env PROVIDER_API_KEY` if you\n  want the config to reference an environment variable instead of storing plaintext. For local\n  Ollama, omit the key.\n\nThe default embedding setup is local `sentence-transformers-local` with\n`intfloat\u002Fmultilingual-e5-small`. Add `--embedding-provider` and `--embedding-model` only when you\nwant to override that default. Use `--dry-run` to preview the files and exec-form commands before\nwriting anything.\n\nFor Codex and Claude Code, native hooks are the default lifecycle path for automatic recall and\nturn capture. Their MCP server uses `--mcp-tools none` by default, so no memX tools are exposed to\nthe agent; this prevents duplicate recall\u002Fwrite and prevents the agent from reading audit data as a\nside channel. Use `--mcp-tools full` only when you intentionally want the agent to see the complete\nMCP tool set. Generic MCP quickstart stays `full` by default because it has no native lifecycle\nhooks. Default native memories are also host-scoped, so Codex and Claude Code do not share the same\nlocal database unless you deliberately override the database path and actor settings.\n\n### Claude Code\n\nThis installs the shared memX config, a local Claude Code plugin marketplace, native lifecycle\nhooks, and the plugin-provided MCP server in one run.\n\n```bash\nnpx -y -p github:NeoLi00\u002FmemX memx quickstart claude-code \\\n  --llm-provider openai-compatible \\\n  --llm-base-url https:\u002F\u002Fllm.example.com\u002Fv1 \\\n  --llm-model fast-memory-model \\\n  --llm-api-key sk-your-provider-key\n```\n\n### Codex\n\nThis installs the shared memX config, Codex MCP config, and native lifecycle hooks in one run.\n\n```bash\nnpx -y -p github:NeoLi00\u002FmemX memx quickstart codex \\\n  --llm-provider openai-compatible \\\n  --llm-base-url https:\u002F\u002Fllm.example.com\u002Fv1 \\\n  --llm-model fast-memory-model \\\n  --llm-api-key sk-your-provider-key\n```\n\n### OpenClaw\n\n```bash\nnpx -y -p github:NeoLi00\u002FmemX memx quickstart openclaw \\\n  --llm-provider openai-compatible \\\n  --llm-base-url https:\u002F\u002Fllm.example.com\u002Fv1 \\\n  --llm-model fast-memory-model \\\n  --llm-api-key sk-your-provider-key\n```\n\n### Generic MCP\n\n```bash\nnpx -y -p github:NeoLi00\u002FmemX memx quickstart mcp \\\n  --llm-provider openai-compatible \\\n  --llm-base-url https:\u002F\u002Fllm.example.com\u002Fv1 \\\n  --llm-model fast-memory-model \\\n  --llm-api-key sk-your-provider-key\n```\n\nFor Claude Code, Codex, and generic MCP clients, start the shared local service after configuration:\n\n```bash\nnpx -y -p github:NeoLi00\u002FmemX memx-server\n```\n\n## Clean uninstall\n\nEach uninstall command backs up the target config first, then removes only memX-owned entries.\nClaude Code and Codex cleanup also uninstall the native plugin, remove the local marketplace, and\ndelete the generated marketplace snapshot.\nOpenClaw cleanup also removes stale `memx` \u002F `memory-memx` slot, allow, and entry references, then\nbest-effort uninstalls both current and legacy plugin files if OpenClaw can still see them.\n\n```bash\nnpx -y -p github:NeoLi00\u002FmemX memx uninstall openclaw\nnpx -y -p github:NeoLi00\u002FmemX memx uninstall codex\nnpx -y -p github:NeoLi00\u002FmemX memx uninstall claude-code\n```\n\nAdd `--dry-run` to preview, or `--config \u002Fpath\u002Fto\u002Fconfig` when using a non-default config path.\n\n## What memX can do\n\n- **Remember work over time**: project decisions, user preferences, task status, long source\n  segments, and raw evidence stay linked to the original turn.\n- **Connect related things**: projects, repos, tools, files, resources, blockers, and outcomes can\n  be represented as entities and graph edges.\n- **Learn collaboration patterns**: repeated evidence can become reusable guidance without losing\n  its supporting sources.\n- **Maintain itself**: corrections can supersede older facts, stable evidence can be promoted, and\n  stale task state stops competing with current state.\n- **Recall compact evidence**: facts, events, state, chunks, relationships, resources, and learned\n  patterns are searched together, then injected as small evidence lines.\n","memX 是一个为AI代理设计的自学习、自维护的记忆系统。它能够将已完成的工作转化为结构化、可搜索且自动维护的记忆，并仅向代理注入当前查询所需的证据。该项目使用TypeScript编写，支持与Codex、Claude Code和OpenClaw等主流AI平台的原生集成，并通过统一的本地记忆层兼容任何MCP客户端。其架构针对长上下文记忆检索进行了优化，在LongMemEval-S基准测试中达到了94.2%的成功率。适用于需要高效管理和利用历史数据以提高决策质量的各种AI应用场景，如工程案例分析等。","2026-06-11 03:55:32","CREATED_QUERY"]