DSH Plugins Marketplace

DSH Plugins

Plugins

/

Memory & Context

/

deepbot

d

deepbot

Identified★ 213

A standalone, always-on agent built on DeepSeek Harness — Slack Socket Mode adapter, in-process sessions, memory-ready gateway profile


DEEPSEEK-BOT

DEEPSEEK-BOT, an agent that remembers: a blue twin-tailed girl with a whale hairclip and a whale laptop, a plush whale on her shoulder, on black with blue pixel blocks

The agentic memory & gateway for DeepSeek Harness. Install once, then just type dsh.

CI GitHub Stars Node License

deepseek-bot is a plugin for DeepSeek Harness.
One always-on agent that remembers across conversations, in your terminal, Slack, Discord and Telegram.
Its commands run in a sandbox, and anything destructive waits for a yes in the conversation.

Install

curl -fsSL https://raw.githubusercontent.com/rlaope/deepseek-bot/master/install.sh | bash
dsh

Messengers, the model provider and other options →


What you get

  • An agent in your terminal.
    dsh is a fullscreen terminal: answers stream in, tool calls show with their diffs.
    It writes docx, xlsx and pptx, reads links with a headless browser, looks at images, and uses MCP tools.

  • Memory that survives.
    Every conversation is a durable session that survives restarts.
    A persona, what it records about you and the work (USER.md, MEMORY.md, memory/), and a recall index
    over past conversations come into each turn. Recall stays inside its own channel; not found is an answer.

  • Where people already talk.
    Slack, Discord and Telegram: mention it and it answers in the thread, 👀 then progress then the end mark.
    stop ends a turn, usage shows tokens, be quiet mutes it with a 🫡 until it is named again.

  • Approvals in the conversation.
    A command that deletes or overwrites waits for the person who asked: Allow or Deny, in place.

  • Contained.
    Commands run under a Seatbelt sandbox that writes only its own workspace and cannot read tokens or credentials.
    Text that tries to give it orders is flagged before the model sees it.

  • DeepSeek's own API by default.
    A key the harness already has is used as it is. OpenGateway or another OpenAI-compatible endpoint is an option.


dsh

Just type dsh, and the agent remembers.

A real session: it reads a file, saves a fact, recalls it in a new conversation, and asks before deleting.
Recorded on the optional OpenGateway route; DeepSeek's own API is the default.

dsh: the agent reads todo.md, saves 'I only deploy on Thursdays' to USER.md, recalls it after /new, and asks before rm; the answer is Deny

[!TIP] Sponsored by OpenGateway, an OpenAI-compatible gateway that puts many model providers behind one API. An option, never a requirement.


Others

Setup · How it works · Limitations · Specification · Plan · Changelog

Install options

It needs macOS with Node.js 22+, the DeepSeek Harness app and a DeepSeek API key; none is asked for if the harness already has one. When the app is missing, the installer offers to download it from DeepSeek's own server, checked against the build's published hash. Running the line again updates; so does dsh update.

dsh gateway setup      # connect Slack, Telegram or Discord: ↑↓ and Enter
dsh gateway status     # is it up, and where it found its key

Without a terminal: installing without a terminal.

Messengers
MessengerVerified
Slackused live
Discordagainst mocked APIs only
Telegramagainst mocked APIs only

On Telegram and Discord anyone who finds the bot can message it: list who may with dsh gateway setup telegram --users <your id>.

What's verified

The unit suites run in CI against fakes; scenarios against a real model run by hand, or daily once service/install-scenarios.sh schedules them; Slack and the terminal are in live use. What is not verified is listed item by item in PLAN.md.

  • macOS only: on Windows the terminal runs by hand, but the installer, services and sandbox are not ported.
  • The sandbox is a policy boundary, not a separate OS user; an MCP server runs outside it, so add only ones you trust.
  • DeepSeek Harness is a release candidate and this project leans on its internals: pin your version.
For contributors and agents

Working rules · Vocabulary

License

MIT. deepseek-bot is an independent project, not affiliated with or endorsed by DeepSeek. The DeepSeek name and logo belong to their owner.

Comments

Loading…

From the same category

hindsight

Hindsight, agent memory that learns: long-term project memory with auto recall and retain, knowledge pages, deep reflection, and per-repo memory banks.

Memory & Context

★ 53.1k

MIT

Python

Index only — not installable

by volcengine

Self-evolving Context Database for AI Agents. Unify Agent Memory, Knowledge RAG and Skills.

Memory & ContextManifest valid

★ 40.7k

↓ 12.9k/wk

AGPL-3.0

Python

Aug 25, 2026

dsh plugin --profile agent add @openviking/dsh-memory-plugin

by Tencent

Open-source LLM knowledge platform: turn raw documents into a queryable RAG, an autonomous reasoning agent, and a self-maintaining Wiki.

Memory & ContextManifest valid

★ 33.2k

↓ 3.7k/wk

NOASSERTION

Go

Aug 25, 2026

dsh plugin --profile web add @wxg-prc-cpg/dsh-weknora

by EverMind-AI

One portable memory layer for every AI agent: local-first, Markdown-native, user-owned, and self-evolving across apps, tools, and workflows.

Memory & Context

★ 13.9k

Apache-2.0

Python

Nov 10, 2026

Index only — not installable

by MemTensor

Self-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.

Memory & ContextManifest valid

★ 12.1k

↓ 3k/wk

Apache-2.0

TypeScript

Aug 25, 2026

dsh plugin --profile web add @memtensor/memos-local-plugin

by plastic-labs

Memory library for building stateful agents

Memory & Context

★ 8k

AGPL-3.0

Python

Nov 10, 2026

Index only — not installable