# DeepSeek Harness Open-source agent harness from DeepSeek with a plugin architecture, web interface and desktop app, currently a developer preview. Canonical profile: https://ki-agentenkompass.de/en/agents/deepseek-harness/ Editor: [Michael Meierhoff](https://ki-agentenkompass.de/en/about/) Sources reviewed: 2026-11-05 Basis: provider documentation. Examples are editorial suggestions, not hands-on product tests. ## Overview DeepSeek Harness is the runtime for agents, not the language model. It runs with DeepSeek models, bundled third parties or custom endpoints. CodeWhale is a separate project that, according to its repository, is not affiliated with DeepSeek. - Provider: DeepSeek - Product family: DeepSeek Harness - Profile type: Open-source agent harness in preview - Environment: Local web interface on 127.0.0.1:3080, desktop app for macOS (Apple silicon) and Windows (64-bit), plus headless, SDK and ACP profiles - Connections: MCP client, plugins under the dsh-plugin topic, ACP and a Python SDK; as model providers DeepSeek, bundled third parties and custom endpoints; GitHub webhook as an optional extension - Setup: Install Node.js 24.9.0 or newer and run `npx @deepseek-ai/dsh web`, or download the desktop app, or build from source with pnpm. Then add model access and choose a workspace. The project advises using a disposable VM, a container or a dedicated environment. - Pricing model: The software is MIT licensed. According to the terms of use, official models need a DeepSeek account and a balance; with your own models you pay their provider. ## Key features ### Everything is a plugin According to the README, the harness is built on a plugin architecture with the Cordis framework. Profiles stack plugin bundles and your own configuration layers. Source: https://github.com/deepseek-ai/deepseek-harness/blob/master/README.md ### Web, headless, SDK and ACP `dsh web` starts the web interface. There are also profiles for one-off headless jobs, SDK clients and ACP. All of them use the same `dsh` command. Source: https://github.com/deepseek-ai/deepseek-harness/blob/master/apps/cli/README.md ### Work in the workspace The agent can read and edit workspace files, run commands, delegate work and maintain a plan. Source: https://github.com/deepseek-ai/deepseek-harness/blob/master/docs/user/guide/index.md ### Subagents and agent teams Subagents run as managed child sessions. The team package adds a shared task board; all members share one checkout. Source: https://github.com/deepseek-ai/deepseek-harness/blob/master/packages/team/agent-team/README.md ### Automation tasks Reminders repeat by an RFC 5545 rule and deliver an instruction to an existing or new conversation. You can create them through the model or by hand. Source: https://github.com/deepseek-ai/deepseek-harness/blob/master/docs/user/guide/schedule.md ### Freely configurable models Besides DeepSeek you can add bundled providers such as anthropic, openai or moonshotai. Custom endpoints speak OpenAI Chat Completions, OpenAI Responses or Anthropic Messages. Source: https://github.com/deepseek-ai/deepseek-harness/blob/master/docs/user/guide/providers.md ### Sandbox and permission presets Permission presets bundle a sandbox mode with an approval policy. The local sandbox stops with an error when it is unavailable instead of running commands unconfined. Source: https://github.com/deepseek-ai/deepseek-harness/blob/master/packages/sandbox/sandbox-local/README.md ## Example workflows ### Assess a repository - Your input: A project folder as the workspace and a request for an overview. - Possible result: The agent reads the code and names the main packages. - What to check: Start in a disposable environment and check the statements against the code. ### Schedule a recurring check - Your input: An automation task with an instruction, a workspace and a repeat rule. - Possible result: The host delivers the instruction to a conversation at the scheduled time. - What to check: Read the result in the conversation: delivery records only confirm the delivery, not the completed work. ## Requirements and limits ### Developer preview without an audit Limitation: SAFETY.md calls the project experimental: no security audit, not to be treated as secure or production-ready. The README announces breaking changes. All 37 releases listed by the GitHub API are pre-releases, most recently dsh-v0.2.2-rc.1 from 4 November 2026. Source: https://github.com/deepseek-ai/deepseek-harness/blob/master/SAFETY.md ### Node.js 24.9.0 or newer Requirement: The repository requires Node.js 24.9.0 or newer, and pnpm for source builds. The product page lists the desktop app only for macOS with Apple silicon and 64-bit Windows. Source: https://github.com/deepseek-ai/deepseek-harness/blob/master/package.json ### Model access required Requirement: According to the terms, official models need a DeepSeek account and a balance. Otherwise you use another provider's model API with your own key. Source: https://deepseek.com/harness/terms-of-use/ ### The sandbox is no guarantee Limitation: According to the terms, the sandbox and approvals lower the risk but do not guarantee isolation. The terms assign responsibility for agent actions to you. Source: https://deepseek.com/harness/terms-of-use/ ### Retention of conversations Not yet confirmed: The privacy policy gives no fixed period for conversations. Network logs are kept for at least six months. Source: https://deepseek.com/harness/privacy/ ## A good fit if … - you want to test an open-source harness locally and extend it with plugins - you want to choose between DeepSeek, third parties and custom endpoints - you can run it in a disposable VM or a container ## Less suitable if … - you need a production-ready, audited solution for sensitive data - breaking changes between versions would be a problem for you - you expect commitments on data location and legal framework for Europe from the provider Editorial assessment based on provider documentation. Not a product test. ## Usable AI models Open choice Besides DeepSeek you choose providers from the installed catalog, such as Anthropic, OpenAI, Kimi or GLM, each with your own key. A custom API also connects gateways and self-hosted servers that speak the OpenAI or Anthropic protocols. The whole project is a developer preview, and the docs do not mention Ollama. - Model families named: DeepSeek, Anthropic (Claude), OpenAI (GPT), Moonshot (Kimi), Z.ai (GLM) · Other models can be connected freely - Model choice: You choose - Your own API key: Yes - Router or custom endpoint: Yes - Local models: Yes Documented · Sources accessed on: 2026-11-05 - [DeepSeek Harness: Providers](https://github.com/deepseek-ai/deepseek-harness/blob/master/docs/user/guide/providers.md) - [DeepSeek Harness: README](https://github.com/deepseek-ai/deepseek-harness/blob/master/README.md) ## Europe & regional availability Read the privacy policy (as of 20 September 2026) and the terms of use. They do not mention the EU, the EEA, the UK or Switzerland. Statements on location only reflect the wording of the policy. Reviewed scope: Official models through DeepSeek and custom models reviewed separately Reviewed on: 2026-11-05 ### Official models: storage in China Availability still unclear The policy says personal data collected in domestic operations is stored in the People's Republic of China and is currently not transferred abroad. Conversations may be used for training after encryption and de-identification; you can opt out through the contact channel. Region covered: EU/EEA, United Kingdom and Switzerland Documented reason: The policy makes no separate statement for users in Europe; availability and legal bases there remain open. Official source: https://deepseek.com/harness/privacy/ ### Custom models: data with the chosen provider With requirements With custom models, according to the policy, DeepSeek Harness does not store or process inputs and outputs on its servers. They go to the provider you configured, whose terms apply. Region covered: EU/EEA, United Kingdom and Switzerland Documented reason: The data path depends on the configured provider and on plugins or MCP servers that process data themselves. Official source: https://deepseek.com/harness/privacy/ ### Legal framework of the terms of use Availability still unclear The terms place the contract under the law of mainland China, with the courts at DeepSeek's domicile. They name no regional restrictions. Region covered: EU/EEA, United Kingdom and Switzerland Documented reason: The reviewed pages do not say whether different European rules apply. Official source: https://deepseek.com/harness/terms-of-use/ ## Accounts & permissions Based on provider documentation. Permissions depend on the account, integration and organization policies. Review the permissions actually requested before use. Configuration covered: Locally run harness; an account only for official models Sources reviewed: 2026-11-05 ### Account used Documented: Official models need a DeepSeek account; the privacy policy names the phone number for registration. Custom models work without an account. The web interface uses a launch token and a signed session cookie. The reviewed pages describe no SSO or roles. Evidence: https://deepseek.com/harness/privacy/ Evidence: https://github.com/deepseek-ai/deepseek-harness/blob/master/docs/user/guide/public-deployments.md ### Admin approval Documented: Permission presets bundle the sandbox and approval policy, such as workspace-write with asking or danger-full-access without asking. The web interface asks before actions that the active policy sends for approval. Evidence: https://github.com/deepseek-ai/deepseek-harness/blob/master/packages/interaction/permission-presets/README.md Evidence: https://github.com/deepseek-ai/deepseek-harness/blob/master/docs/user/guide/index.md ### Data access & team sharing Documented with conditions: According to SAFETY.md, the project can run code and commands, load third-party plugins and access the network, processes, credentials and files made available to it. How far this reaches depends on the sandbox and the environment. Evidence: https://github.com/deepseek-ai/deepseek-harness/blob/master/SAFETY.md ## Prices and included usage Pricing checked: 2026-11-05 The product page names no price. The software is MIT licensed; costs arise with the model provider. DeepSeek publishes an API price list. ### DeepSeek Harness Free under the MIT license Software with web interface and plugins; model use not included. Price source: https://github.com/deepseek-ai/deepseek-harness/blob/master/README.md ### deepseek-flash (API) Input without cache $0.15 to $0.3, output $0.6 to $1.2 per 1M tokens (off-peak to peak) Model version DeepSeek-V4.1-Flash-1017; cache hits $0.003 to $0.006. Price source: https://api-docs.deepseek.com/quick_start/pricing ### deepseek-v4-pro (API) Input without cache $0.66 to $1.32, output $1.98 to $3.96 per 1M tokens (off-peak to peak) Model version DeepSeek-V4-Pro-0813; cache hits $0.022 to $0.044. Price source: https://api-docs.deepseek.com/quick_start/pricing The API price list uses the $ sign without a currency code. The reviewed pages do not say whether official models on a Harness balance are billed exactly this way. Third-party pricing was not used as evidence. Pricing details: https://api-docs.deepseek.com/quick_start/pricing ## Operation & oversight Based on public provider documentation. Assessments are editorial inferences from the stated properties. Actual time requirements and reliability have not been measured. Deployment covered: Self-run process on your computer or server Sources reviewed: 2026-11-05 ### Files and working state Documented: The process uses its start directory as the default location; the interface needs a selected workspace. Sessions live in an append-only event log. According to its README, the base bundle provides durable sessions that survive restarts. Evidence: https://github.com/deepseek-ai/deepseek-harness/blob/master/docs/user/guide/index.md Evidence: https://github.com/deepseek-ai/deepseek-harness/blob/master/packages/core/session/README.md Evidence: https://github.com/deepseek-ai/deepseek-harness/blob/master/packages/bundle/base/README.md ### Sign-ins Documented: API keys live in `$DSH_HOME/.credentials.yaml`, and the settings only hold a reference. The page does not show saved keys again. The printed start URL contains a process token and is meant only for intended users. Evidence: https://github.com/deepseek-ai/deepseek-harness/blob/master/docs/user/guide/providers.md Evidence: https://github.com/deepseek-ai/deepseek-harness/blob/master/docs/user/guide/public-deployments.md ### Starting further runs Documented with conditions: Automation tasks start by rule while the host runs. An optional webhook overlay creates a review session when a GitHub pull request becomes ready. If the host is closed, scheduling stops. Evidence: https://github.com/deepseek-ai/deepseek-harness/blob/master/docs/user/guide/schedule.md Evidence: https://github.com/deepseek-ai/deepseek-harness/blob/master/docs/user/guide/github-review.md ### Human intervention Documented: The web interface asks when the approval policy requires it. The terms and SAFETY.md recommend least privilege and human approval for significant actions. Evidence: https://github.com/deepseek-ai/deepseek-harness/blob/master/docs/user/guide/index.md Evidence: https://deepseek.com/harness/terms-of-use/ Evidence: https://github.com/deepseek-ai/deepseek-harness/blob/master/SAFETY.md ### Detecting errors Documented with conditions: The local sandbox reports SANDBOX_UNAVAILABLE and separates failures from policy denials. Delivery records only confirm delivery, not the model work. The reviewed pages describe no central error view for users. Evidence: https://github.com/deepseek-ai/deepseek-harness/blob/master/packages/sandbox/sandbox-local/README.md Evidence: https://github.com/deepseek-ai/deepseek-harness/blob/master/docs/user/guide/schedule.md ### Resuming after failures Documented with conditions: Saved sessions can be resumed. After a context overflow, the basic package condenses and retries. After a restart, the host sends an overdue one-shot task or the latest missed occurrence; exactly-once delivery is not guaranteed. Evidence: https://github.com/deepseek-ai/deepseek-harness/blob/master/packages/core/agent-loop/README.md Evidence: https://github.com/deepseek-ai/deepseek-harness/blob/master/packages/compaction/compaction-basic/README.md Evidence: https://github.com/deepseek-ai/deepseek-harness/blob/master/docs/user/guide/schedule.md ### Editorial assessment The harness describes the sandbox, approvals and sessions at length in the repository. It remains a preview without an audit, though. You run it yourself and, according to the terms, bear the consequences of its actions. ## Memory and conversation continuity Scope: No built-in memory; optional MCP servers Sources accessed: 2026-11-05 The shipped composition includes no memory server. Three MCP examples are off by default; storage and retrieval depend on the chosen system. ### Storage and scope Partly documented: The shipped composition has no memory server. The examples store data where the third-party system decides, such as `~/.memorix/data`, `~/.engram` or a JSONL file in the home directory. - [DeepSeek Harness: Connect a third-party memory MCP server](https://github.com/deepseek-ai/deepseek-harness/blob/master/docs/user/guide/mcp-memory.md) ### Selection and timing Partly documented: DSH makes no selection of its own. The docs optionally suggest an instruction: on request the model writes through a memory tool. Selection rules belong to the third-party system. - [DeepSeek Harness: Connect a third-party memory MCP server](https://github.com/deepseek-ai/deepseek-harness/blob/master/docs/user/guide/mcp-memory.md) ### Inspect, edit and delete Not publicly established: The page describes no interface of its own for review, editing or deletion. This depends on the chosen server. - [DeepSeek Harness: Connect a third-party memory MCP server](https://github.com/deepseek-ai/deepseek-harness/blob/master/docs/user/guide/mcp-memory.md) ### Agent and channel boundaries Partly documented: According to the page, Memorix and Engram derive the project from the Git project in the DSH working directory. The page does not describe boundaries between agents or channels. - [DeepSeek Harness: Connect a third-party memory MCP server](https://github.com/deepseek-ai/deepseek-harness/blob/master/docs/user/guide/mcp-memory.md) ### Retrieval Partly documented: Retrieval runs through the server's MCP tools. The reference server searches by substring, not semantically. - [DeepSeek Harness: Connect a third-party memory MCP server](https://github.com/deepseek-ai/deepseek-harness/blob/master/docs/user/guide/mcp-memory.md) ### Session reset and compaction Partly documented: Compaction summarizes older history and also runs through `/compact`; the original stays in the session log. The package docs describe no link to memory. - [DeepSeek Harness: dsh-compaction-basic](https://github.com/deepseek-ai/deepseek-harness/blob/master/packages/compaction/compaction-basic/README.md) - [DeepSeek Harness: dsh-session](https://github.com/deepseek-ai/deepseek-harness/blob/master/packages/core/session/README.md) ## Context Pre-release status without a security audit, not cleared for production use. Breaking changes are announced. Node.js 24.9.0 or newer for the npx start. The terms and the privacy policy do not mention the EU, the EEA, the UK or Switzerland. Read the repository (README, SAFETY.md, docs, package READMEs, release list via the GitHub API), the product page, the privacy policy, the terms of use and the API price list. Feature claims come from the provider; example tasks are editorial. Our own hands-on tests are pending. ## Open points Europe, availability still unclear: - Official models: storage in China - Legal framework of the terms of use Requirements and limits, not yet confirmed: - Retention of conversations ## Sources - https://deepseek.com/harness - https://github.com/deepseek-ai/deepseek-harness - https://github.com/deepseek-ai/deepseek-harness/releases - https://github.com/deepseek-ai/deepseek-harness/blob/master/package.json - https://github.com/deepseek-ai/deepseek-harness/blob/master/docs/architecture.md - https://github.com/deepseek-ai/deepseek-harness/blob/master/apps/cli/README.md - https://deepseek.com/harness/terms-of-use/ - https://deepseek.com/harness/privacy/ - https://github.com/deepseek-ai/deepseek-harness/blob/master/README.md - https://github.com/deepseek-ai/deepseek-harness/blob/master/docs/user/guide/index.md - https://github.com/deepseek-ai/deepseek-harness/blob/master/packages/team/agent-team/README.md - https://github.com/deepseek-ai/deepseek-harness/blob/master/docs/user/guide/schedule.md - https://github.com/deepseek-ai/deepseek-harness/blob/master/docs/user/guide/providers.md - https://github.com/deepseek-ai/deepseek-harness/blob/master/packages/sandbox/sandbox-local/README.md - https://github.com/deepseek-ai/deepseek-harness/blob/master/SAFETY.md - https://api-docs.deepseek.com/quick_start/pricing - https://github.com/deepseek-ai/deepseek-harness/blob/master/docs/user/guide/public-deployments.md - https://github.com/deepseek-ai/deepseek-harness/blob/master/packages/interaction/permission-presets/README.md - https://github.com/deepseek-ai/deepseek-harness/blob/master/packages/core/session/README.md - https://github.com/deepseek-ai/deepseek-harness/blob/master/packages/bundle/base/README.md - https://github.com/deepseek-ai/deepseek-harness/blob/master/docs/user/guide/github-review.md - https://github.com/deepseek-ai/deepseek-harness/blob/master/packages/core/agent-loop/README.md - https://github.com/deepseek-ai/deepseek-harness/blob/master/packages/compaction/compaction-basic/README.md - https://github.com/deepseek-ai/deepseek-harness/blob/master/docs/user/guide/mcp-memory.md ## Provider According to the terms and privacy policy, the company behind the product is Hangzhou DeepSeek Artificial Intelligence Basic Technology Research Co., Ltd. It also develops the DeepSeek language models. - Operator: DeepSeek - Organization: Provider - Website: https://deepseek.com - Reviewed: 2026-11-05 - Provider source: https://deepseek.com/harness - Provider source: https://deepseek.com/harness/terms-of-use/ - Provider source: https://deepseek.com/harness/privacy/ [Sources and method](https://ki-agentenkompass.de/en/sources/)