dsh-graft-plugin
A DeepSeek Harness plugin that puts graft — a prebuilt graph of every symbol, its file:line span, and who calls what — in front of both you and the model.
Install
dsh plugin --profile web add dsh-graft-plugin
dsh --profile web
The install puts the graft CLI on your PATH as well. Then index a repo once:
cd <your repo>
graft build
That is the whole setup. The chip appears in the composer and the model gets the tools.
What it does
Names the graph you are actually on
A chip beside the model name, resolved from the session’s workspace rather than from wherever dsh was launched:
● graft · my-repo in sync
◐ graft · my-repo syncing
▲ graft · my-repo stale

This is the reason the plugin exists rather than the graft mcp server. graft mcp resolves its repo once, from its own working directory at startup, and nothing can re-point it afterwards. Launched from a home directory it answers “no graph found” for every workspace you open — while the CLI, run in the same repo, answers fine.
Gives the model five tools
| tool | what it answers |
|---|---|
graft_ask |
“how does X work” / “where is Y” — ranked hits, code inlined at exact file:line |
graft_grep |
every occurrence, grouped by enclosing symbol |
graft_callers |
who calls it, what it calls, the blast radius before a rename |
graft_skeleton |
a file’s whole API in ~200 tokens |
graft_map |
orientation in an unfamiliar tree |
Each is resolved per session, so one dsh can serve several workspaces at once.
None of them take a boolean parameter — they take string enums, which reads better to a model and avoids a real failure mode on some OpenAI-compatible servers where a call that sets a boolean is emitted as literal <tool_call> text and never runs.
Keeps the index current
/graft builds or rebuilds on demand. After the model edits files the index rebuilds in the background, debounced so a long turn does not rebuild constantly. It reuses graft’s own sync runner, so it cannot drift from what graft build does.
Draws the graph
A graft viz button opens the visualiser as a tab inside dsh-better-sidebar, if you have it mounted.

The code view drills into a single file — what it contains, what it depends on, what depends on it:

Configuration
On the plugin’s row in your profile’s cordis.patch.yml — all optional:
- id: graft-status
name: 'dsh-graft-plugin'
config:
cwd: 'C:/path/to/repo' # pin to one repo instead of resolving per session
tools: false # drop the graft_* tools, keep the chip
command: false # drop /graft
promptSection: false # stop telling the model to prefer graft
autoBuild: false # stop rebuilding after edits
Requirements
- dsh
0.1.2-rc.1or newer, web profile - Node 22+
- the
graftCLI on PATH — installed for you, ornpm install -g @nanonets/graft
Licensed MIT. Issues and pull requests at github.com/adithyanraj03/dsh-graft-plugin.