MIT-licensed native DeepSeek Harness plugin for Graphify, the local code knowledge graph engine: build and query a graph of symbols, dependencies and architectural communities with `/graphify`, with a compact mode for smaller local models.
DSH integration
Native runtime
Author-claimed
Safety audit
Unaudited
Last verified
2026-10-31
License
MIT
01What can it help you accomplish?
Answer architectural questions about a repository from a graph-backed map instead of repeated path-guessing and grepping
Symbols, dependency paths, architectural communities and graph-backed project context drawn from `graphify-out/graph.json`
Developers running DeepSeek Harness sessions on a repository they are exploring or onboarding onto
Assess the blast radius of a change before touching an interface
Dependency paths to callers and dependents, shortest paths, owning communities and pull-request impact analysis (full mode)
Engineers reviewing risky changes inside a DSH conversation
Keep smaller local models consistently using the graph during a session
Compact mode with six high-frequency tools: status, graph query, node inspection, neighbours, shortest path, and the session-scoped graph resource
Teams running local 20B–30B models in DeepSeek Harness where tool-selection capacity is limited
02How to install into DeepSeek Harness
Prerequisites
- A working DeepSeek Harness (DSH) profile
- Node.js 22.18 or later
- Python 3.10 or later
- `uv` — needed for the Graphify runtime install
Installation steps
- 01
Add the plugin to your DSH profile: run `dsh plugin --profile <profile> add dsh-graphify` (replace <profile> with the profile you normally use; the bundle activates itself, no Cordis hand-editing needed)
$ dsh plugin --profile <profile> add dsh-graphify
- 02
Install the local Graphify runtime: run `uv tool install 'graphifyy[mcp]==0.9.57'` — if you install it after DSH has started, ask for `graphify_status` and the plugin discovers it on the next check or tool call, without a restart
- 03
Build a graph: start DSH from the repository you want to explore (`cd /path/to/your/repository && dsh --profile <profile>`) and run `/graphify` in the conversation — a full build that writes `graphify-out/graph.json`
Verify the integration
- Run `graphify_status` in the DSH conversation: it reports the resolved project, graph path, runtime availability, connection state, and whether the graph matches the indexed source state
03DSH integration and capability boundaries
Native DSH plugin added via `dsh plugin add` — resolves the active session project, manages the Graphify MCP process, and registers graph tools in the conversation
/graphify graph build and update
the repository the DSH session was started from→a full build writing `graphify-out/graph.json`, or a conservative code-only incremental update via `/graphify update`
writes `graphify-out/graph.json` in the repositoryincremental updates run only for changes the plugin can prove are code-only; otherwise the graph stays marked stale until a full `/graphify` buildFull and compact tool modes
`toolMode` config — `full` (default) or `compact`→full exposes the complete Graphify capability set including communities, god nodes and PR impact; compact registers six focused navigation tools for local 20B–30B models
graphify_status diagnostics
asked for in the DSH conversation→resolved project, graph path, runtime availability, connection state, and whether the graph matches the indexed source state
Session-aware project resolution and isolation
the active DSH workspace→graph requests follow the active session project
model-invoked access to projects outside the active workspace is blocked by default; `allowExternalProjects: true` opts in deliberatelyResilient local runtime management
the locally installed Graphify runtime (uv tool)→automatic runtime discovery, reconnection after unexpected MCP-process exits, optional `graphifyVersion` pinning for the uv fallback
the plugin manages the Graphify MCP process for the session
04Who is it for? When not to use it?
Good for
- Developers running DeepSeek Harness sessions on a repository they are exploring or onboarding onto
- Engineers reviewing risky changes inside a DSH conversation
- Teams running local 20B–30B models in DeepSeek Harness where tool-selection capacity is limited
Not for
- Incremental updates cover code extraction only. After documentation, package metadata or configuration changes the graph goes stale until a full `/graphify` build; `freshness.mode` defaults to `warn` (set `auto` for safe pre-query updates, `off` only for a deliberately static graph).
05Compatibility, maintenance and safety notes
- Tested on macOS and Ubuntu Linux only — Windows is not yet verified. The stack also needs Node.js 22.18+, Python 3.10+ and `uv`.
- Supports `@deepseek-ai/cordis` >=4.0.4 <5; tested against DeepSeek Harness 0.2.0-rc.2 (and the 0.2.1-alpha channel) and Graphify 0.9.57, with contract and schema-drift checks also passing on Graphify 0.9.90.
- Incremental updates cover code extraction only. After documentation, package metadata or configuration changes the graph goes stale until a full `/graphify` build; `freshness.mode` defaults to `warn` (set `auto` for safe pre-query updates, `off` only for a deliberately static graph).
- Graph requests follow the active DSH workspace; model-invoked access beyond it is blocked unless you deliberately set `allowExternalProjects: true`.
- BENCHMARK.md defines a planned evaluation protocol for Graphify-assisted agents — it is an experimental protocol, not published performance results.
MIT · actively maintained (latest release v0.2.1, published 2026-10-13)
06Frequently asked questions
What is Graphify?
Graphify is a local graph engine for code: it builds a knowledge graph of a repository's symbols, dependencies and architectural communities. The upstream project lives at Graphify-Labs/graphify on GitHub. dsh-graphify is the DeepSeek Harness-native plugin layer that connects a DSH session to that graph — engine and plugin are installed separately (`uv tool install 'graphifyy[mcp]==0.9.57'` for the engine, `dsh plugin add dsh-graphify` for the plugin).
How do I install dsh-graphify?
Three steps: add the plugin to your DSH profile with `dsh plugin --profile <profile> add dsh-graphify`; install the runtime with `uv tool install 'graphifyy[mcp]==0.9.57'`; then start DSH from the repository you want to explore and run `/graphify`. Prerequisites: a working DSH profile, Node.js 22.18+, Python 3.10+ and `uv`.
How does Graphify turn a codebase into a knowledge graph?
Run `/graphify` in a DSH session started from your repository: the command performs a full build and writes `graphify-out/graph.json`, mapping symbols, dependencies, architectural communities and god nodes. Updates are conservative — incremental updates cover code extraction only, so run a full build after documentation or configuration changes.
What is the difference between full and compact mode?
The default `full` mode exposes the complete Graphify capability set, including communities, god nodes and pull-request impact. `compact` registers six high-frequency tools — status, graph query, node inspection, neighbours, shortest path and the session-scoped graph resource — and is recommended for local 20B–30B models, where fewer choices help smaller models use the graph consistently.
Does dsh-graphify work on Windows?
The plugin is tested on macOS and Ubuntu Linux; Windows is not yet verified. It supports `@deepseek-ai/cordis` >=4.0.4 <5 and is tested against DeepSeek Harness 0.2.0-rc.2 (plus the 0.2.1-alpha channel) with Graphify 0.9.57 — schema-drift checks also pass on Graphify 0.9.90.
07Related DSH workflows
weknora
by tencent
Open-source LLM knowledge platform: turn raw documents into a queryable RAG, an autonomous reasoning agent, and a self-maintaining Wiki.
honcho
by plastic-labs
Memory library for building stateful agents
webnovel-writer
by lingfengqaq
A long-form Chinese web-novel writing assistant for Claude Code and DeepSeek Harness, designed to reduce memory loss and hallucinations in serialized projects of up to 2 million characters.
mirage
by strukto-ai
The World's First Virtual Terminal for AI Agents
08Data and sources
`dsh-graphify` connects [Graphify](https://github.com/Graphify-Labs/graphify)'s code knowledge graph to DeepSeek Harness…
The plugin resolves the active session project, manages the Graphify process, and provides `graphify_status` for diagnos…
This page is generated from the project’s public documentation, repository metadata and a structured parse of DSH Plugins; last verified on 2026-10-31. Found an error? Submit a correction.
Best DeepSeek Harness Plugins
Twelve plugins worth installing first — picked from the whole catalog, across every category.