Apache-2.0, local-first memory kit that turns agent conversations and resources into an editable, searchable, interconnected Markdown knowledge base — and ships a native DeepSeek Harness plugin that recalls memory before each step and auto-captures turns.
DSH integration
Native runtime
Author-claimed
Safety audit
Unaudited
Last verified
2026-08-25
License
Apache-2.0
01What can it help you accomplish?
Give the DeepSeek Harness agent a local, file-based long-term memory layer
ReMe Markdown memory recalled before each agent step, with completed main-agent turns auto-captured into the knowledge base
DeepSeek Harness (dsh) users who want persistent, user-editable memory that carries across sessions
Turn agent conversations and external resources into a searchable, interconnected Markdown knowledge base
Daily memory cards and long-term digest nodes (personal / procedure / wiki) linked by wikilinks
Agents and developers who want an inspectable, backup-able memory workspace they fully control
02How to install into DeepSeek Harness
Prerequisites
- ReMe requires Python 3.11+.
- Basic file operations, BM25 search, wikilink traversal, and reading proactive topics can run without LLM credentials.
Installation steps
- 01
Install ReMe: `pip install "reme-ai[core]"`
- 02
Start the ReMe service: `reme start` (default address 127.0.0.1:2333)
- 03
Add the plugin to the DeepSeek Harness Web profile: `dsh plugin --profile web add @agentscope-ai/reme`
$ dsh plugin --profile web add @agentscope-ai/reme
Verify the integration
- `reme health_check` reports the service status
- `curl -s http://127.0.0.1:2333/version` returns the running version
- `reme version` prints the installed ReMe version
03DSH integration and capability boundaries
Native DeepSeek Harness plugin: start the ReMe service, then `dsh plugin --profile web add @agentscope-ai/reme`; the plugin recalls ReMe memory before each agent step and auto-captures completed turns.
Self-evolving memory capture (Auto Memory / Auto Resource / Auto Dream)
filtered conversation source records, agent sessions, and external resources→daily memory cards and long-term digest nodes (personal / procedure / wiki) with wikilinks
Progressive hybrid memory search
a natural-language query with optional line-range and link-expansion hints→matching chunks with line ranges and bounded wikilink neighbors, fused via RRF
Agent-friendly local integration (CLI / HTTP / MCP / Python)
host process or external agent via CLI, HTTP API, MCP server, or Python API→read / write / maintain / reuse of the same local Markdown workspace
Proactive interest discovery
digest nodes produced by Auto Dream→structured interest topics the host agent may choose to act on
04Who is it for? When not to use it?
Good for
- DeepSeek Harness (dsh) users who want persistent, user-editable memory that carries across sessions
- Agents and developers who want an inspectable, backup-able memory workspace they fully control
05Compatibility, maintenance and safety notes
- `proactive` only reads and exposes interest topics from Auto Dream; it does not browse the web, send notifications, or rewrite the knowledge base — the host agent decides whether and how to act.
- Semantic embedding retrieval is disabled by default; enabling it requires uncommenting `components.as_embedding` and `components.embedding_store` in reme/config/default.yaml and providing an embedding API key.
Apache-2.0 · actively maintained (latest release v0.4.1.7, 2026-08-13)
06Frequently asked questions
How do I add ReMe to DeepSeek Harness?
Start the ReMe service (`reme start`), then run `dsh plugin --profile web add @agentscope-ai/reme` in your DeepSeek Harness Web profile. The plugin recalls relevant memory before each agent step and submits finished turns for automatic capture.
Does ReMe need an embedding model or API key to work?
No. Embeddings are disabled by default, so basic file operations, BM25 search, wikilink traversal, and reading proactive topics run without LLM credentials. LLM_API_KEY is only required for auto_memory, auto_resource, and auto_dream.
What does ReMe store, and who controls it?
ReMe treats memory as plain Markdown files (frontmatter + wikilinks) under `.reme/` in your working directory. Users and agents can inspect, edit, move, and back up the files directly; everything under metadata/ is rebuildable.
How is the DeepSeek Harness integration different from the MCP / HTTP integrations?
The `@agentscope-ai/reme` npm package is a native DeepSeek Harness plugin installed via `dsh plugin`. Other agents connect through the streamable HTTP MCP service, the CLI/SKILL.md, or the embedded Python API.
Is my data sent to the cloud?
ReMe is local-first. By default it runs at 127.0.0.1:2333 and only calls external LLM/embedding endpoints if you configure LLM_API_KEY / EMBEDDING_API_KEY; without them, file operations and BM25 search stay fully local.
07Related DSH workflows
weknora
by tencent
Open-source LLM knowledge platform: turn raw documents into a queryable RAG, an autonomous reasoning agent, and a self-maintaining Wiki.
mirage
by strukto-ai
The World's First Unified Virtual Filesystem For AI Agents
graph-memory
by adoresever
Knowledge-graph memory core native to DSH (Cordis adapter, cross-session triples).
memtrace-public
by syncable-dev
Structural memory for AI coding agents. Bi-temporal graph, MCP-native, zero LLM calls. Cursor · Claude Code · Codex · DeepSeek Harness · Hermes · VS Code · Windsurf.
08Data and sources
providing a native ReMe memory integration for DeepSeek Harness.
dsh plugin --profile web add @agentscope-ai/reme
The plugin recalls relevant ReMe memory before agent steps and submits completed main-agent turns for automatic memory c…
This page is generated from the project’s public documentation, repository metadata and a structured parse of DSH Plugins; last verified on 2026-08-25. Found an error? Submit a correction.
Best DeepSeek Harness Plugins
Twelve plugins worth installing first — picked from the whole catalog, across every category.