DeepSeek Harness: Open-Source AI Agent Harness + Desktop App

DeepSeek Harness (DSH) is DeepSeek’s open-source AI agent harness for working with code, documents, and research in a local workspace. Ask it to investigate a repository or prepare a report, and it can read files, run commands, search for information, and record the steps it takes.

DSH runs on Cordis, where the model connection, tools, skills, sessions, and even the interface are built from plugins. You can extend the agent with new capabilities or configure it for different kinds of work through the same system.

DeepSeek Harness ships as a desktop app for macOS and Windows, alongside a local Web UI, CLI, and Python SDK. In the desktop app, you can open a project, follow the agent’s actions, and manage the plugins that shape its work.

Key Features

• Official desktop app plus a local Web UI.
• Local file editing, shell commands, code changes, document work, and Web research.
• Persistent sessions with resume, fork, search, replay, and Trajectory details.
• Background jobs and built-in Web automation tasks for reminders and recurring work.
• Standard, PTC, Minimal, and Creator Agent Presets for different tool configurations.
• Plugin discovery, installation, updates, and agent-assisted extension creation.
• CLI, headless, ACP, and Python SDK entry points for developer use.

Everything Is a Plugin

Cordis mounts plugins, resolves their dependencies, and connects the services and events they expose. DeepSeek Harness uses this architecture across the major parts of an agent, and Profiles determine which plugin bundles load for a particular setup.

100 Best DeepSeek Harness Plugins for DSH: Find community plugins for memory, terminals, search, browser control, UI extensions, and agent features.

Install, Update, and Create Plugins

You can manage official bundles and third-party extensions from Plugins. Enabling, disabling, uninstalling, and updating plugins synchronizes their dependencies. With hot reload enabled, replaceable plugins can update during a running session.

For additional community sources, enable the experimental Custom plugin indexes bundle in Plugins. Open Discover -> Add index source, enter the URL of a plugin index, and select a version to install.

To create a custom plugin, select Let Agent create a plugin to open Creator mode with the current draft preserved. Describe the plugin you need, such as a custom tool or UI panel, and the agent begins creating plugin files after you send the request.

Use the Official Desktop App

The official DeepSeek Harness desktop app is available for Apple Silicon Macs and 64-bit Windows. It works with local files and keeps active tasks running in the background after you close the main window.

It can open workspace files in local apps, manage plugins, search the model list, handle application updates, and run Windows sandbox permission diagnostics.

Run the Web UI from the CLI

Launch the Web UI

Install Node.js and open a terminal in the project directory you want DSH to access. A local launch opens the default browser and serves the Web UI at http://127.0.0.1:3080.

Pass --no-open when only the server should start. Over SSH, dsh prints the host URL and leaves local forwarding to the SSH client or editor.

npx @deepseek-ai/dsh web

Connect a Model

Open Settings -> Models and enter a DeepSeek API key, or use Add model provider for a bundled third-party provider or a custom API endpoint.

Choose the Workspace

Click Choose workspace, register the project directory, and select it. The directory that launched dsh is the default filesystem location, and the session composer becomes available after a workspace is selected.

Run Your First Task

Summarize this repository and identify its main packages.

Run It from Source

git clone https://github.com/deepseek-ai/deepseek-harness.git
cd deepseek-harness
pnpm install
pnpm run build
pnpm dsh web

Use Files, Documents, and Images

• The Web UI can preview Markdown, code, HTML, PDF, and image files in its file sidebar.
• Preview Word, Excel, PowerPoint, CSV, and TSV files.
• Local files can open in an installed editor, terminal, file manager, or associated desktop app.
• Uploads continue in the background with progress and cancellation controls.
• Image input depends on the active model. Your models can enable Image under Settings -> Models -> Model options -> Input types.

Sessions, Trajectory, and Background Jobs

DeepSeek Harness keeps an append-only session history that records prompts, reasoning events, tool calls, tool results, delegated work, and injected context. The Trajectory view exposes the run behind the visible conversation.

Persistent sessions can be resumed, forked, searched, and replayed. Use /compact in the Web or desktop composer to summarize older context while continuing a long conversation. Archived-session controls manage older work, and background jobs keep long commands or delegated tasks running outside the immediate chat turn.

The Web app also includes Automation tasks for scheduled reminders and recurring work. Reminder tools are available to Standard, Creator, and PTC agents.

Profiles, Agent Presets, and Subagents

A Profile determines which plugin bundles load when dsh starts. DSH currently includes Profiles for Web, headless, SDK, minimal SDK, and ACP use, and custom Profiles can define their own plugin bundle set and configuration.

Agent Presets

The Web and desktop apps include four built-in Agent Presets, which control the tools and behavior available to a new task.

Claude Code, Codex, and Other Subagents

You can use Claude Code and Codex as optional official bundles that add external subagent providers. Install them from the official plugin catalog as needed, then configure an Agent Preset to delegate independent tasks. These bundles can also be used with headless, SDK, and ACP Profiles.

Connect DeepSeek or Another Model Provider

You can use DeepSeek Harness with DeepSeek’s official models, third-party APIs, or a custom model endpoint. DeepSeek account sign-in is available in the desktop app and the local Web UI. If you already have a DeepSeek API key, add it under Settings → Models. The model picker lists all the models available through each connection, and the API-key catalog updates as DeepSeek publishes changes.

For Anthropic, OpenAI, Moonshot AI (Kimi), and Z.ai (GLM), choose Add model provider and enter your credentials. DSH supplies the endpoint settings and model catalog for these built-in providers. A private gateway or another model service can be added through Custom model API, using its base URL, model IDs, and API protocol.

DSH accepts OpenAI Chat Completions, OpenAI Responses, and Anthropic Messages.

Select a model in the conversation composer to make it the default for new sessions. Existing sessions retain the model recorded in their history, and changes to model settings take effect on the next request.

DSH stores API keys locally in $DSH_HOME/.credentials.yaml and displays only redacted credentials in Settings. Headless runs read credentials from environment variables or .env files by default.

Use DeepSeek Harness from Python

The Python SDK is the programmatic entry point to the Harness runtime. It requires Python 3.10 or newer, Git, a DeepSeek endpoint and credential, and an isolated workspace and Harness home.

The runtime packages are available for Linux x64, Linux arm64, macOS 14 or newer on Apple silicon, and Windows x64.

python -m pip install deepseek-harness-sdk

Set DEEPSEEK_API_KEY in your process environment, then create a DeepSeekHarness instance with an explicit workspace and Harness home. The SDK starts its bundled runtime, runs the task in a named session, and returns the final response.

from deepseek_harness import DeepSeekHarness
with DeepSeekHarness(
    provider="deepseek-official",
    model="deepseek-v4-flash",
    max_tokens=49152,
    cwd="/absolute/path/to/workspace",
    dsh_home="/absolute/path/to/dsh-home",
    profile="sdk-minimal",
) as harness:
    result = harness.run(
        "Summarize the repository and identify its main packages.",
        session_id="example-001",
    )
print(result.final_response)

The Python and TypeScript SDKs can resume saved sessions with their original IDs and working directories. The SDK also exposes file delivery, questions, scheduling, skills, context, question callbacks, and operation-approval controls for applications that integrate the Harness runtime.

Developer Preview and Safety

DeepSeek Harness is currently in developer preview, and releases can introduce compatibility-breaking changes.

The runtime can execute model-generated commands and third-party plugin code against the files, processes, network connections, and credentials available to it.

Sandboxing, approval prompts, and permission controls restrict supported operations but cannot guarantee isolation.

Use a restricted workspace and limited privileges for tasks that execute untrusted code.

Pros

• Official desktop app plus Web, CLI, and Python entry points.
• Plugin architecture extends models, tools, skills, UI, jobs, and agent behavior.
• Plugin manager installs extensions, and Creator mode can build plugins and custom presets.
• Persistent sessions, Trajectory view, subagents, and background jobs.
• DeepSeek, bundled third-party providers, and custom endpoints.

Cons

• Preview releases can include compatibility-breaking changes.
• Third-party plugins can execute commands with the runtime’s permissions.

Alternatives & Related Resources

• 100 Best DeepSeek Harness Plugins for DSH
• CodeWhale (DeepSeek TUI): Free, Open-source Claude Code Alternative
• Reasonix: Free DeepSeek Coding Agent for CLI, Desktop & VS Code
• Pi Coding Agent: Extensible AI Terminal Agent for Developers
• Kimi Code CLI: Open-Source AI Coding Agent with Skills & MCP Support
• Automate Anything: 10 Best & Open-source AI Agents
• DeepSeek Timeline: Model Release Dates and Key Milestones
• Vercel fx: Open-Source Native AI Coding Agent in Zig

FAQs

Is DeepSeek Harness free to use?

The MIT-licensed Harness is free to install. DeepSeek’s official model service requires prepaid credit through the DeepSeek Open Platform, and other API providers charge under their own pricing. A locally hosted model uses your own computing resources.

Can DeepSeek Harness use a model running on my computer?

Yes. Run the model through a local server that exposes OpenAI Chat Completions, OpenAI Responses, or Anthropic Messages, and add its address under Custom model API. DSH runs the agent and its tools, while the local model server handles inference.

Does DeepSeek Harness send my project files to a model provider?

DSH stores session history locally and reads workspace files through local tools. When a task uses file contents, tool results, or other workspace context in a model request, that information is sent to your selected model endpoint. For custom-model connections, DeepSeek states that inputs and outputs go directly to the chosen provider and are not stored on DeepSeek Harness servers.

Is there an official DeepSeek Harness desktop app for Linux?

There is no official Linux desktop download. DeepSeek provides desktop apps for Apple silicon Macs and 64-bit Windows. Linux users can run the Web UI or CLI, or use the Python SDK on x64 and arm64 systems.

Can I install Claude Code plugins in DeepSeek Harness?

Standard Claude Code plugins cannot be installed directly through DSH’s plugin manager. DSH uses Cordis plugins, and its official Claude Code and Codex bundles add those tools as delegated agents. An experimental bridge supports parts of the Claude Code Mods API, but it does not load standard Claude Code plugin manifests.

Last Updated: November 1, 2026