toolclub/dsh-agent-team-gui
Persistent multi-model squads configured in Settings and selected in the Composer, with per-member model and tool policies. By default the lead Agent completes simple work itself and delegates only when the saved squad is needed; delegated work runs as a bounded DAG with optional review and repair. Run Center tracks retries and per-member tokens from official Harness usage events.
Install
via GitHub · GitHub source
Installing from GitHub runs the project's build script, which pnpm blocks until you allowlist it — run the command once and pnpm prints the exact key to add under `allowBuilds` in ~/.dsh/profiles/web/pnpm-workspace.yaml.
What happened when we ran it
Installs, but needs a build approved first
pnpm refuses to run a package's build script until it is allowlisted. Add the key pnpm prints to allowBuilds in the profile's pnpm-workspace.yaml, then run the command again.
Every command here is run in a throwaway container against a clean profile, and the result is whatever the harness recorded — not a guess from the source. Last run 15d ago.
Show it in your README
For maintainers: the badge serves this listing's latest sandbox verdict, so a re-run updates it on its own — and it links readers to the full result here.
Due diligence
What it is — Configure persistent multi-model teams supporting member model and tool policies.
Who it is for — Users who need multi-model collaboration for complex tasks using planner and reviewer models are suitable for this. Users who do not require dynamic workflow planning can skip this plugin.
Watch out — Sandbox test shows it can be installed but pnpm blocks the build script for github:toolclub/dsh-agent-team-gui. Users need to manually allow the build to register properly. No other obvious issues found.
The verdict — If I were to install it, I would use the compiled release package to avoid build issues, but still need to manually allow the build script for source installation.
Generated by grok-4.6, and a starting point rather than a verdict. Where it says a plugin installs or does not, that is from a real run in a clean profile — everything else is read off the repository. Trust the source over this.
dsh-agent-team-gui is a DeepSeek Harness plugin that manages persistent agent teams and runs them from the normal conversation composer. DeepSeek Harness loads the plugin in the Web profile, where it adds Teams and Members settings plus Team, Solo, and Inherited conversation modes. DeepSeek Harness uses the selected conversation's provider/model route for a bounded, tool-free planner when dynamic planning is enabled. The planner reads member roles and returns structured assignments and an acyclic dependency graph; ready members then run concurrently up to maxConcurrency, while dependent members receive bounded structured handoffs. Fixed order instead runs members serially. Team runs write durable history containing plans, stages, outputs, errors, timings, retries, review or repair rounds, and official token-usage coverage.
Each member can have its own model route, role prompt, fallback route, token limit, and tool policy. The plugin supports Always, Smart, and Manual activation, all-member or adaptive selection, foreground or background work, and optional resilience, budget, and quality-gate controls. Members can be granted terminal tools, so the terminal surface reaches whatever terminal operations those policies permit; review permissions before running a team.
dsh-agent-team-gui suits developers who repeatedly coordinate planners, implementers, reviewers, or specialists across conversations. It is a poor fit when a normal single-agent send is sufficient, or when the required DeepSeek Harness Web profile, supported Node.js version, pnpm, and configured provider/model route are unavailable. Node.js 23 is unsupported, and Git installation may require authorising the package's prepare build.
DeepSeek Harness loads dsh-agent-team-gui in the Web profile. Open Settings → Members to define reusable members with a provider/model route, role prompt, optional fallback route, token limit, and permitted tools. In Settings → Teams, combine members and choose:
Fixed order;Always, Smart, or Manual activation;All members or Adaptive selection;The conversation control beside the normal composer supports Team, Solo, and Inherited. A project default can be followed through Inherited; a queued Team or Solo selection applies to the next eligible message only.
Dynamic planning uses the active conversation's provider/model route in a bounded, tool-free planner child. The planner receives member roles and returns structured assignments with an acyclic dependency graph. Invalid, cyclic, or unavailable plans fall back to deterministic role-scoped assignments. Ready nodes run up to maxConcurrency; dependent members receive bounded structured handoffs, while complete output remains in durable run history. Fixed order bypasses DAG planning. A quality gate can request up to two repairs from a named repair owner.
The Team runs view exposes plans, stages, members, review or repair rounds, outputs, errors, timings, retries, and official token coverage. Conversation states remain available after refresh, Host startup, and live reconnect.
Use dsh --profile web --dump-config to inspect the composed configuration; its output should contain the dsh-agent-team-gui bundle layer and an agent-team-gui row. Git dependencies run the repository's prepare build. With pnpm 10 or later, authorise only this package by adding dsh-agent-team-gui: true under allowBuilds in ~/.dsh/profiles/web/pnpm-workspace.yaml, then restart the running DeepSeek Harness Web process after installation or updates.
DeepSeek Harness must be >=0.1.0-rc.5 <0.2.0, using the Web profile. The environment needs Node.js >=22.19.0 <23 or >=24.0.0; Node.js 23 is unsupported, and pnpm plus at least one configured DSH provider/model route are required. A compiled release tarball does not need Git prepare permission.
Written from the project's own documentation and kept in sync with it. Where the two disagree, the source is authoritative — read the README on GitHub
Same category
Q00
Config-only bundle that mounts Ouroboros through the DSH MCP client, exposing 36 interview, Seed, execution, evaluation, and evolution workflow tools in DSH.
No one-line install — this plugin lives inside a larger repository and publishes no npm package.
loopx-project
LoopX, a provider-neutral, local-first state kernel and control plane for long-horizon agents: keeps Goal, Todo, gate, evidence, quota, recovery, and handoff state above DeepSeek Harness, while the plugin bootstraps the CLI and skills, admits bounded same-session continuation, and adds a loopback GoalBar for the exact bound loop.
No one-line install — this plugin lives inside a larger repository and publishes no npm package.
huangruiteng
LoopX, a provider-neutral, local-first state kernel and control plane for long-horizon agents: keeps Goal, Todo, gate, evidence, quota, recovery, and handoff state above DeepSeek Harness, while the plugin bootstraps the CLI and skills, admits bounded same-session continuation, and adds a loopback GoalBar for the exact bound loop.
No one-line install — this plugin lives inside a larger repository and publishes no npm package.
chuspeeism
Embeds the active installed Codex Taskboard runtime in the DeepSeek Harness sidebar, using its launcher runtime descriptor instead of a fixed port.
No one-line install — this plugin lives inside a larger repository and publishes no npm package.
tong-io
TongFlow film-crew studio for image, voice, music and video production: the agent writes per-asset TongFlow workflow files (.tongflow.json) that run through TongFlow plugins, with an embedded workflow canvas, a shot/character/take project layout and a manga-drama template; sessions starting with @tongflow open the Studio view.
No one-line install — this plugin lives inside a larger repository and publishes no npm package.