DSH Plugins Marketplace

DSH Plugins

Plugins

/

agent-memory-hub

z

agent-memory-hub

Discovered

Agent memory bridge — local-first, human-approved, file-native memory for Claude Code, Codex, opencode & gemini

agent-memory-hub

Agent memory bridge — local-first, human-approved, file-native long-term memory for AI coding agents.

One local service. Every agent shares it. Claude Code, Codex, opencode, gemini-cli, DSH — anything that speaks MCP or a shell.

It is

  • Local-first — your memory lives in plain Markdown files you own (knowledge/), versioned in history/, git-friendly. No cloud required. No black-box database.
  • Human-approved — agents can propose memories; only you can commit them. Every stored fact is traceable: who approved it, when, and from which conversation.
  • File-native — Markdown + YAML frontmatter + JSONL ledger. Open it, diff it, grep it, back it up with git.
  • Cross-host by design — one MCP server + one CLI (memhub) + one optional local dashboard, all backed by the same files. Register once, every agent shares the same memory.
  • Scoped — project → domain → global knowledge hierarchy with ancestor-chain search. Each agent sees exactly the layers it should.

Install

DSH plugin (DeepSeek Harness users):

install_bundle dsh-agent-memory-hub

Tool surface: 8 memory_* tools (search / get / status / propose / preflight / cite / journal / close-task). Runtime is zero-network — a dead remote never degrades your host. Details: host/dsh/README.md · 部署对接指南.

CLI / MCP server (Claude Code, OpenCode, Codex, gemini-cli, any MCP host):

npm i -g agent-memory-hub
memhub mcp --root <hub>       # register in any MCP client
memhub extract material       # scheduled-extraction material for any workflow platform
memhub publish                # publish approved entries to your team's memory-hub API (ADR-0015)

Team memory-hub service: implement or deploy docs/02-spec/memory-hub-api-v1.md (reference server included: examples/memory-hub-server.mjs).

Status

🚧 Under active development (this repo was founded 2026-11-15). The stable, working edition of the same memory format is dsh-memory-hub (DSH plugin, v0.8.x). agent-memory-hub generalizes it into a standalone cross-host service; the first stable release targets 0.1.0.

Design principles

Four invariants (see AGENTS.md):

  1. Files are the authority, not a cache.
  2. Humans approve; agents propose. Never self-commit.
  3. Plain text only — readable, diffable, migratable.
  4. Zero network by default.

Eight day-zero architecture decisions in docs/01-architecture/decisions/ADR-0001-day-zero-principles.md.

Roadmap

StepDeliverable
1 ✅Repo skeleton, invariants, format contract tests
2core — storage, retrieval (BM25), workspace pyramid registry
3transport — MCP stdio + CLI, re-verified on four hosts
4Approval + ledger + local dashboard (memhub serve)
5Injector / extractor first-class commands (memhub pack / hook / extract)

License

MIT

Comments

Loading…

Similar plugins

memsearch

by zilliztech

A persistent, unified memory layer for all your AI agents (e.g. Claude Code, Codex, DSH), backed by Markdown and Milvus.

Memory & ContextManifest valid

★ 2.9k

↓ 12.2k/wk

MIT

Python

Nov 13, 2026

dsh plugin --profile agent add @zilliz/memsearch-dsh

Read-only bridge that surfaces Claude Code's existing project memory in DSH — git-root-aware resolution, credential redaction before injection, and an on-demand recall tool. Zero runtime dependencies.

Memory & ContextManifest valid

★ 0

↓ 143/wk

dsh plugin --profile web add dsh-claude-memory

by 398894496-arch

Second brain for coding agents. Seal the day, distill into Obsidian, hit that page tomorrow. Cursor, Codex, Claude Code, DeepSeek Harness.

Memory & ContextManifest valid

★ 39

MIT

Python

Oct 11, 2026

dsh plugin --profile web add dsh-krouter

by tinqiao-oss

A portable memory protocol for AI agents — load it as standing rules; a curation discipline + reference spec + optional cap hook.

Memory & ContextManifest valid

★ 207

↓ 1.6k/wk

MIT

Python

Nov 17, 2026

dsh plugin --profile agent add dsh-engramory

by jonah791

Agent-driven long-term memory for DeepSeek Harness: scoped memory (global + per-workspace), layered entries (fact/knowle

Manifest valid

★ 3

↓ 188/wk

MIT

TypeScript

Sep 7, 2026

dsh plugin --profile web add dsh-agent-memory

by annkun

Deepseek harness memory

Memory & ContextManifest valid

★ 4

↓ 664/wk

MIT

TypeScript

Oct 20, 2026

dsh plugin --profile web add @fooxe/dsh-memory