DS Bot Brings Persistent AI Agent Teams to DeepSeek Harness at 4,600 Stars
DS Bot is a multi-agent plugin for DeepSeek Harness, now at 4.6k GitHub stars. Ships as a desktop app, DSH plugin, or Docker image; runs fully local under Apache 2.0. Bots have persistent memory, per-bot model selection, and can be configured by chatting with Chief.
Takeaways
- DS Bot is a multi-agent plugin for DeepSeek Harness, now at 4.6k GitHub stars.
- Ships as a desktop app, DSH plugin, or Docker image; runs fully local under Apache 2.0.
- Bots have persistent memory, per-bot model selection, and can be configured by chatting with Chief.
- Group chats feature an admin bot, @mention handoffs between bots, and isolated group memory.
- Deep Office support: create and edit .docx, .pptx, .xlsx with tracked changes and slide operations.
- Built on MCP for capabilities; local commands require user approval before execution.
DS Bot , an open-source plugin for DeepSeek Harness , has crossed 4,600 GitHub stars while still at version 0.2 preview. It turns the harness into a team of long-lived bots that can work in direct messages or group chats, retain memory, and delegate tasks through @mentions.
DeepSeek Harness, or DSH, manages model sessions, tools, and provider configuration. DS Bot adds a desktop interface and collaboration layer built around specialized roles such as writer, coder, and chief of staff. A main bot named Chief administers the team, coordinates group work, and guides users through the initial setup.
One app, three deployment paths
| Option | Best suited to | What it includes |
|---|---|---|
| Desktop app | Windows, macOS, and Linux users | Bundles DSH, so no separate harness installation is required. |
| DSH plugin | Existing DeepSeek Harness setups | Installs through the repository’s single npx command. |
| Docker image | NAS devices and always-on servers | Provides a container boundary and persistent deployment. |
The application, orchestration layer, and saved state run locally, with persistent data stored under ~/.dsh . Model inference runs through the provider configured in DSH, so prompts and files may leave the machine when a remote endpoint is used. Token usage and billing also remain tied to that provider.
Memory and handoffs shape the workflow
| Feature | Behavior |
|---|---|
| Persistent bot memory | Each bot retains preferences and working context for later conversations. Memory can be edited from the bot details page or through chat. |
Reporting is indexed from AlphaSignal. Rights remain with the original publisher and cited sources.
Read original report
