MemTensor/MemOS
C3Self-evolving memory OS for LLM & AI Agents: ultra-persistent memory, hybrid-retrieval, and cross-task skill reuse, with 35.24% token savings and DeepSeek Harness support.
★ 11k+ · MemTensor/MemOS source on GitHub · this plugin in the registry
MemTensor/MemOS is a DeepSeek Harness plugin rated C3 — powerful capability combined with sensitive behavior. It patches the dsh runtime, runs code at install time, can rewrite the system prompt, executes system commands.
Installable plugin — declares a dsh.bundle manifest
What it can do
| Capability | Flag | Evidence |
|---|---|---|
| patches the dsh runtime | runtime_patch | ./adapters/deepseek-harness/cordis.patch.yml |
| runs code at install time | install_script | postinstall: electron-rebuild && npm --prefix skills/dev-browser install && npm --prefix skills/file-permission install && npm --prefix skills/ask-user-question install |
| can rewrite the system prompt | prompt_surface | systemPrompt |
| executes system commands | exec | ×18 in authored code, e.g. apps/memos-local-openclaw/scripts/postinstall.cjs:4, apps/memos-local-openclaw/scripts/postinstall.cjs:4 |
| decodes base64 payloads | base64_decode | ×7 in authored code, e.g. apps/memos-local-openclaw/src/hub/auth.ts:22, apps/memos-local-plugin/core/hub/auth.ts:65 |
| starts a network server | net_server | ×4 in authored code, e.g. apps/memos-local-plugin/server/http.ts:87, apps/memos-local-openclaw/src/hub/server.ts:71 |
| reads credential-class env vars | token_env | EMBEDDING_API_KEY, OPENCODE_GO_API_KEY, OPENCODE_ZEN_API_KEY, SUMMARIZER_API_KEY |
Services it injects
llm systemPrompt tools
Hooks it attaches
agent/pre-step llm/adapters-updated session/disposed session/event
Outbound domains
api.openai.com generativelanguage.googleapis.com api.anthropic.com api.mistral.ai api.voyageai.com memos.memtensor.cn api.cohere.ai fonts.googleapis.com api.cohere.com memos-dashboard.openmem.net
Environment variables it reads
USERPROFILE MEMOS_ARMS_ENDPOINT OPENCLAW_STATE_DIR MEMOS_ARMS_PID MEMOS_ARMS_ENV OPENCLAW_CONFIG_PATH EMBEDDING_ENDPOINT SUMMARIZER_ENDPOINT EMBEDDING_API_KEY EMBEDDING_MODEL
How to read this
Levels measure capability surface and transparency, not maliciousness. A C3 plugin can be entirely legitimate — a desktop shell genuinely needs subprocesses. The point is that you can see this before installing. See the levels explained and how dsh plugins work.
Findings come from static analysis of shipped code; nothing is executed. Think a flag is wrong? Open an issue — every flag cites the file and line it came from.