Plugins

Connect DeepSeek Harness to 3Chat via OAuth for customer and conversation search, context retrieval, confirmed messaging, and batch-send tracking.

Shares a local memory pool between DeepSeek web chat and Harness, with full-answer saving and sidebar browsing and injection. The optional Windows desktop patch for DSH 0.2.2-rc.1 adds shadow sessions that sync completed web turns to Harness, display Harness replies in the original web chat, and hand tasks off to Harness.

Drive your local DSH from WeChat: a self-hosted iLink bridge over a persistent ACP session (context lives in DSH and is resumable), phone conversations grouped into the desktop workspace, with image recognition, file messages (Excel/Word/PDF, parsed by the agent itself) and voice transcripts. Tool approvals are answered inside WeChat itself (numbered card; ask/allow/reject policy; 120s timeout means reject), and any process can push a message to your phone through the outbox helper. Access defaults to strict (only registered devices are served); the phone channel's permission preset is configurable - see the README security section before sharing the bot.

Model switching and reasoning-effort control for DeepSeek Harness: set_model and list_models agent tools let the agent autonomously change its provider, model or thinking depth (reasoning effort) mid-task with a mandatory reason, a provider whitelist and a context-window token-pressure guard, the active model is injected into the runtime context, and Plan Mode auto-transition switches to a configured planning model on entering plan mode and restores the previous execution model on exit.

Takes over the edit/write rows in the chat stream with the official native row shell and collapse behavior; expanding shows per-file code cards with file-type icon, add/delete stats, per-block collapse, highlight/plus-minus toggle, syntax-highlighted unified diff with real file line numbers, and a configurable context margin. Also ships a patch tool (beta, off by default) that applies multi-spot multi-file patches atomically via the *** Begin Patch envelope.

NeatMem long-term memory plugin for DeepSeek Harness — thin REST client against a neatmem server

Unofficial DeepSeek Harness integration for Unreal Engine 5.8 Unreal MCP

Lossless context folding for DeepSeek Harness: the oldest span is replaced by a brief, and the originals stay in the session log.

Live skill provider that adds skills from 60+ other agents' directories (Claude Code, Codex, Cursor, Gemini CLI, Copilot), Claude Code plugin marketplaces, and any git repository of skills to the dsh catalog without copying or symlinking; deduplicates, repairs frontmatter, renames name collisions, and ships a CLI (list, add, sync, doctor).

Installs a Web-profile helper that diagnoses and writes MCP config so Codex, Cursor, or Claude Code can spawn dsh --profile codex.

Prompt enhancement for the composer: an "✨ Enhance" button rewrites your draft prompt into a clearer, more structured version with first-principles thinking, grounded in session context, memory and recent conversation; asks clarifying questions in a card when key info is missing, then previews the result before applying it to the input.

Codex-style review of session file changes: per-turn cards (optionally taking over the shell edited cards) whose totals are the turn net diff, a review tab with syntax-highlighted diffs that never wrap, real line numbers, open-at-line and a draggable split, a hover preview of the whole file with three lines of context around each change, one-click per-turn revert, and the on-disk record size with a clear button in the review toolbar. Records are sharded per session, with content-addressed snapshot blobs that dedupe and are garbage collected. Light/dark presets follow the DSH theme, typography follows DSH and installed skins (or any font on the machine, with a px stepper), and a settings switch injects a think-in-Simplified-Chinese rule into the global AGENTS.md.

Config-driven lifecycle hooks, including context injection: declare event, handler and outcome in YAML, where the handler is an in-harness tool (MCP included), a shell command or an HTTP endpoint, and the outcome is model-visible context, a denied tool call, or fire-and-forget.

A slim port of Headroom: compresses tool output, cuts up to 20% of context, without affecting the model's context cache or agent performance.

Lazy-loading router for MCP servers: auto-discovers compatible MCPs and keeps only a shared router in the tool catalog, revealing a server's schemas when a turn needs them and hiding them again at turn end, so many installed MCPs stop costing context on every request. Falls open to normal visibility on any uncertainty, and the original MCP keeps ownership of execution, permissions and process lifecycle. Also supports explicit lazy connections with warm reuse, bounded reconnection, stdio and Streamable HTTP.

Four-module session companion: smart conversation export (Markdown/PDF/JSON/PNG long-image with privacy redaction and batch ZIP), context handoff summaries with template save/import, API cost optimization (live official pricing, peak/off-peak scheduling, daily/monthly budgets, model routing), and global conversation search with in-chat find (Ctrl+F, CSS Custom Highlight API).

LLM-free lossless* compaction engine for DeepSeek Harness

An always-on working style for DSH agents, packaged as an agent preset: restate the task, read before editing, verify each change with a real command, keep context lean. Ships 29 skills (plan loops, review loops, TDD, unsticking, PR watching) plus a turn guard that pushes back when a turn edits files without a successful check afterwards, and pushes back harder on a second miss.

Exposes a running DeepSeek Harness as an MCP server over Streamable HTTP or a bundled stdio bridge, so an external agent can create workspaces and sessions, prompt, fork, rename and switch the model on them, read and write files, run foreground and background shell commands, call the instance's own agent tools by name, and read per-turn and cumulative token usage with a cost receipt when a price plugin is installed.

Publishes the running DSH host as a Model Context Protocol endpoint, so other harnesses such as Claude Code, Cursor, Codex or another DSH can start and continue sessions, read transcripts, wait on or cancel turns, and answer pending approvals; every call is attributed to its caller, and owner-scoped sessions keep callers apart.

Long-term memory for dsh via a local Hindsight service: each turn is retained with local timestamps and relevant memories (time-aware, reranked) are recalled into context before replies; each workspace binds to one memory bank, several workspaces can share a bank, and banks can have Chinese names; what gets retained is configurable (default: user messages and AI replies); history is imported only when you choose to; includes a three-step setup page with one-click local deployment (uv + Python 3.12), a sidebar status card and a panel to browse and search memories. Needs a SiliconFlow API key for the default embedding and reranker models; only Linux is verified so far.

Takes over the compaction threshold on every mounted engine: forces headroomTokens to -1e6 so the ratio is the only term in the trigger, keeps retention below it on every write path, reroutes the compaction call to a chosen model, and adds a panel with a one-click compaction button and a per-session compaction log.

Windows Computer Use for DeepSeek Harness with 21 tools and verified targeting: controls are addressed through their UI Automation identity and patterns, and numbered marks are checked against the live window before input, so a stale target is refused instead of clicked. A known sequence runs in one computer_batch call, one model round trip; incremental observations, compact receipts (normally no more than about 6 KiB extra), bounded handoff waits (HANDOFF_WAIT_TIMEOUT/HANDOFF_HELD_TIMEOUT as continuations), the desktop-block cycle boundary and a deterministic 100 s owner-scoped leave backstop keep context small and prevent a stuck handoff from waiting out the tool deadline. Approval modes settle an unanswerable ask honestly: in a danger-full-access session with no approval card the action runs under that grant, every other sandbox — and every delegated subagent session — refuses and names the unavailable channel instead of reporting a user rejection, and gate diagnostics land in policy.jsonl. Human handoff, a model-visible stop brake and prebuilt native workers.

Hierarchical Claude Code-style memory: user + project scopes (project memories committed to git for the team), MEMORY.md index auto-injected at session start, context-window guards, pure files.