Apache-2.0, local-first control plane that governs long-horizon agents on DeepSeek Harness (dsh) with durable goal/todo/gate/evidence state and quota-gated, human-in-the-loop turns.
DSH integration
Compatible
Author-claimed
Safety audit
Unaudited
Last verified
2026-09-02
License
Apache-2.0
01What can it help you accomplish?
Run long-horizon, governed agent loops on DeepSeek Harness (dsh) with durable state across bounded turns
Headless dsh segments through LoopX Turn, each tick gated by `quota should-run`, with durable goal, todo, gate, evidence, and quota state preserved across turns and harnesses
Engineers and teams running long-running or multi-agent work on DeepSeek Harness who need reviewable, restartable control instead of chat-only memory
Turn a public issue into a reviewable, evidence-backed change
Reviewable, evidence-backed PR delivery plus reusable, revision-stamped fix knowledge
Open-source contributors and maintainers who want long-running issue-fix work to stay aligned and reviewable
Coordinate multi-agent research (proposer, executor, evaluator/promoter) with visible quota and evidence
Reproducible multi-agent research workflows where todos, quota, evidence, and targeted wake remain visible
Researchers and engineers running reproducible experiments who need hypotheses, evidence, and promote/stop gates visible in one graph
02How to install into DeepSeek Harness
Prerequisites
- Python 3.11+ (the README lists this as the requirement).
- An active Python environment whose console scripts are on `PATH`.
- No runtime dependencies outside the Python standard library.
Installation steps
- 01
Install from PyPI: `python3 -m pip install --upgrade loopx`.
- 02
Install the workflow skills and run the doctor check: `loopx workflow-skills --install` then `loopx doctor`.
- 03
From your project root, connect and check status: `loopx connect` then `loopx status`.
- 04
For DeepSeek Harness: install `loopx[deepseek-harness]`, prepare a dsh `cordis.yml`, then use the dsh goal-mode adapter with `loopx turn run-once`.
Verify the integration
- `loopx doctor` passing;
- `.loopx/registry.json` and a projected active goal state;
Rollback
- The Installing LoopX guide documents upgrade, rollback, uninstall, and the archive fallback.
03DSH integration and capability boundaries
Connects to DeepSeek Harness (dsh) through the dedicated dsh goal-mode adapter: install the `loopx[deepseek-harness]` extra, prepare a dsh `cordis.yml`, and run `loopx turn run-once`; each headless dsh segment is gated by `quota should-run`.
Durable goal/todo/gate/evidence state kernel
a long-horizon objective and bounded agent turns→stable goal, todos, gates, evidence, quota, and handoff state across turns, tools, and agents
Quota-gated, human-in-the-loop turns
a registered-agent tick→a decision to deliver, ask, wait, self-repair, or stay quiet, gated by `quota should-run`
DeepSeek Harness (dsh) goal-mode adapter
a dsh `cordis.yml` and `loopx turn run-once`→headless dsh segments through LoopX Turn, each tick gated by `quota should-run`
04Who is it for? When not to use it?
Good for
- Engineers and teams running long-running or multi-agent work on DeepSeek Harness who need reviewable, restartable control instead of chat-only memory
- Open-source contributors and maintainers who want long-running issue-fix work to stay aligned and reviewable
- Researchers and engineers running reproducible experiments who need hypotheses, evidence, and promote/stop gates visible in one graph
Not for
- LoopX is not an autonomous production controller. Dangerous permissions, publishing, production writes, and final ownership stay with the human.
- LoopX does not grant credentials, approve destructive or production actions, publish on a user's behalf without authorization, or turn an unverified run into evidence of success.
05Compatibility, maintenance and safety notes
- LoopX is not an autonomous production controller. Dangerous permissions, publishing, production writes, and final ownership stay with the human.
- LoopX does not grant credentials, approve destructive or production actions, publish on a user's behalf without authorization, or turn an unverified run into evidence of success.
- LoopX is not a full agent platform, an agent runtime, or an autonomous production controller.
Apache-2.0 · actively maintained (latest release v0.5.3, 2026-08-27)
06Frequently asked questions
How do I connect LoopX to DeepSeek Harness (dsh)?
Install the `loopx[deepseek-harness]` extra, prepare a dsh `cordis.yml`, then use the dsh goal-mode adapter with `loopx turn run-once`. Each headless dsh segment runs through LoopX Turn and is gated by `quota should-run`.
Does LoopX replace DeepSeek Harness or run as its own agent runtime?
No. LoopX runs on top of harnesses rather than replacing them, and it is not another agent framework or a provider-specific orchestration runtime. The harness still executes the bounded turns; LoopX governs the durable state.
What Python version does LoopX require?
Python 3.11+. Use an active Python environment whose console scripts are on `PATH`; the package has no runtime dependencies outside the standard library.
Is it safe to run LoopX unattended on DeepSeek Harness?
LoopX is not an autonomous production controller. Dangerous permissions, publishing, production writes, and final ownership stay with the human, and it will not grant credentials or approve destructive or production actions without authorization.
How do I verify a successful LoopX + dsh connection?
A successful connection has `loopx doctor` passing and shows `.loopx/registry.json` with a projected active goal state; `loopx status` should display the current objective, a concrete user gate, and the next agent todo.
07Related DSH workflows
archify
by tt-a1i
Agent skill for beautiful, verifiable architecture, workflow, sequence, data-flow, and lifecycle diagrams—self-contained HTML with motion and crisp export.
openviking
by volcengine
Self-evolving Context Database for AI Agents. Unify Agent Memory, Knowledge RAG and Skills.
nocobase
by nocobase
NocoBase is an open-source AI + no-code platform for building business systems fast. Instead of generating everything from scratch, AI works on top of production-proven infrastructure and a WYSIWYG no-code interface, so you get both speed and reliability.
learn-harness-engineering
by walkinglabs
Harness engineering beginner tutorial, from 0 to 1
08Data and sources
Runs on top of any agent harness — Codex App, Claude Code, Cursor, dsh, or your own — providing long-horizon state, sema…
Install `loopx[deepseek-harness]`, prepare a dsh `cordis.yml`, then use the [dsh goal-mode adapter](loopx/dsh_goal_mode/…
Headless dsh segments through LoopX Turn, each tick gated by `quota should-run`
This page is generated from the project’s public documentation, repository metadata and a structured parse of DSH Plugins; last verified on 2026-09-02. Found an error? Submit a correction.
Best DeepSeek Harness Plugins
Twelve plugins worth installing first — picked from the whole catalog, across every category.