powercontext-dsh: A Long-Term Memory and Context Plugin for DeepSeek Harness

PowerContext for DeepSeek Harness is a memory and context plugin (DSH plugin) for DeepSeek Harness that connects to a running PowerContext Server over HTTP for recall, memory, handoff, experience, and skills. It does not embed storage, start the Server, or import the Python package; instead, it calls the Server's /v1/... OpenAPI surface. Therefore, both the PowerContext Server and DeepSeek Harness must be running as two separate processes. This plugin solves the problem of long-term memory and context management for agents by automatically recalling bounded context and injecting it as untrusted historical evidence before each model step, while also capturing the current user input as a content source.

What core features does the PowerContext for DeepSeek Harness DSH plugin support?

The plugin calls the Server's /v1/... OpenAPI surface over HTTP (not using MCP). Before each model step, it automatically performs two actions: recalling bounded context with POST /v1/context/prepare and injecting it as untrusted historical evidence, and capturing the current user input as a Content Source with POST /v1/sources/content. Named pc_* tools expose agent-safe Memory, handoff, experience, skill, and read-only review operations, such as pc_search, pc_remember, pc_handoff_activate, pc_experience_generate, pc_skill_get, etc. DSH requests one-time user approval before named mutations run. Review mutations remain explicit human /pc review commands; destructive and administrative OpenAPI operations are not model tools. If the Server is unreachable, recall is skipped and the turn continues.

How to install PowerContext for DeepSeek Harness?

Before installation, install DeepSeek Harness and ensure the web profile exists (run dsh web once). The recommended way is to use the PowerContext CLI's setup dsh command, which automatically calls dsh plugin --profile web add and keeps the ref aligned. You can also install from a GitHub Release tarball of this standalone repository. Installation commands are as follows:

# Using PowerContext CLI (recommended)
powercontext setup dsh --source oceanbase/powercontext --ref master

# Or from a local checkout
powercontext setup dsh --source /path/to/powercontext

# Or manually add the plugin directory
dsh plugin --profile web add /path/to/powercontext/integrations/dsh/plugins/powercontext

# Or install from a tarball
dsh plugin --profile web add ./powercontext-dsh-0.0.3.tgz

If the plugin is already installed from a source checkout, remove it first before installing the tarball (on Windows, replacing a link: install with a tarball fails). After TypeScript changes, rebuild with pnpm install, pnpm test, pnpm build, then restart dsh web. Optional checks: powercontext doctor checks the Server, powercontext doctor dsh checks the CLI and plugin id.

Which DeepSeek Harness versions and platforms are compatible with PowerContext for DeepSeek Harness?

This plugin is maintained by knqiufan, licensed under Apache-2.0, written in TypeScript, and last updated in 2026-08. It depends on several official DSH packages (such as @deepseek-ai/dsh-agent, @deepseek-ai/dsh-tools, etc.) and is also integrated into the official PowerContext repository at integrations/dsh/plugins/powercontext, staying in sync with this standalone repository. Since the plugin communicates with the Server over HTTP, it is theoretically compatible with any platform that supports HTTP, but both the PowerContext Server and DeepSeek Harness must be running.

View more DSH plugins at dsh-plugin.org