{"slug": "tencent-loopforge-self-hosted-sops-that-orchestrate-claude-code-codex-and", "title": "Tencent LoopForge: Self-Hosted SOPs That Orchestrate Claude Code, Codex and CodeBuddy", "summary": "Tencent published LoopForge, an MIT-licensed, self-hosted control plane for orchestrating coding agents, which trended on GitHub with roughly 700+ stars and 174 stars gained in a day as of November 14, 2026. LoopForge runs standard operating procedures (SOPs) that chain agent roles, shell scripts, and human approval gates inside Docker sandboxes, supporting ten executors including Claude Code, Codex, and CodeBuddy, and installs via ./start.sh pulling a pinned ghcr.io/tencent/loopforge image, listening on 0.0.0.0:8090 by default. The release targets team-scale delivery — published workflows, shared requirements, and token-cost reports — rather than single-developer CLI plugins.", "body_md": "**Tencent published [LoopForge](https://github.com/Tencent/LoopForge?ref=explainx) as a self-hosted control plane for coding-agent delivery** — chat with Main Bot for exploratory work, or run **SOPs** (standard operating procedures) that chain agent roles, shell scripts, and human gates in Docker sandboxes. The repository trended with **roughly 700+ stars** and **174 stars gained in a day** as of November 14, 2026, per GitHub counters.\n\nLoopForge sits in the same problem space as [Plannotator’s local review surface](https://explainx.ai/blog/plannotator-plan-code-review-workspace-coding-agents-claude-code-codex-github-trending-2026) and [Agent Autoloop’s background hub](https://explainx.ai/blog/agent-autoloop-claude-code-codex-background-hub-cron-github-issues-approval-cards-github-trending-2026), but Tencent’s angle is **team-scale delivery**: published workflows, shared requirements, token-cost reports, and a web console — not only a single-developer CLI plugin.\n\n## TL;DR\n\n| Question | Answer (per Tencent README, Nov 2026) | \n|---|---|\n| License | MIT | \n| Default install | `./start.sh` pulls pinned`ghcr.io/tencent/loopforge` image | \n| Agent CLIs | Ten executors including Claude Code, Codex, CodeBuddy | \n| Sandboxes | Docker containers with repository checkouts | \n| Human gates | Pause runs until approval or send work upstream | \n| Local edition | `npx loopforge-cli@latest install` for Claude, Codex, CodeBuddy, Cursor | \n| GitHub | Optional OAuth for PR delivery; not required for document SOPs | \n\n## Why a coding agent still needs a delivery layer\n\nAsking an agent to “fix this bug” once is easy. Asking the **same team** to ship **the same way every time** — plan, implement, test, review, approve — is not. LoopForge’s README frames the gap explicitly: demos skip tests, agents review their own diffs, and nobody notices when a run stalls.\n\nLoopForge adds **roles** (planner, implementer, reviewer, tester), **scripts** for deterministic steps, and **human gates** where a person signs off or sends work back to an earlier node. That matches how enterprises already run [Claude for manufacturing plugins](https://explainx.ai/blog/claude-for-manufacturing-plugin-17-skills-infor-netsuite-autodesk-fusion-approvals-2026) and [AWS Bedrock agent routing](https://explainx.ai/blog/aws-bedrock-litellm-nemo-switchyard-coding-agent-routing-2026), but here the orchestration is **yours** on **your Docker host**.\n\n## Two paths: chat with Main Bot or run an SOP\n\n| Mode | Best for | How it runs | \n|---|---|---|\n| Main Bot chat | Exploration, quick fixes, steering mid-reply | Opens a conversation workspace (Docker sandbox) on demand | \n| SOP | Repeatable delivery with fixed stages | Engine runs Agent, Script, Human-gate, Sub-SOP, and Preview nodes in order | \n\nBoth paths attach to the same **requirements** object, so whether code ships from chat or a SOP, history and cost roll up together — closer to ITSM than a one-off `/loop` in the terminal.\n\nHuman approval gates pause agent delivery until a teammate signs off\n\nSOP nodes, per the README:\n\n| Node | Role | \n|---|---|\n| Agent | A configured role with prompt, executor, model, rules, Skills, MCP tools | \n| Script | Shell build, test, lint, or report generation | \n| Human gate | Blocks until someone approves or rejects upstream | \n| Sub-SOP | Nested workflow | \n| Preview build | Static site archive for review | \n\nFailed checks can **rework** to earlier steps — the designer shows this with the built-in bug-fix SOP.\n\n## Install and first run\n\nTencent documents a **Compose-first** path that does **not** require Go or Node on the host:\n\n```\ngit clone https://github.com/Tencent/LoopForge.git\ncd LoopForge\n./start.sh\n```\n\n`./start.sh` checks Docker, Compose, Python 3, and curl, pulls the release image pinned in `compose.yaml`, seeds built-in sandboxes and starter SOPs, and listens on **0.0.0.0:8090** by default. The **first registered account becomes administrator**.\n\nTry without any model key: open the **demo project**, run the **Platform check · script & approval** SOP, approve the human gate, and watch a script-only path complete — useful for proving the engine before you paste API keys.\n\nFor air-gapped or slow `ghcr.io` regions, the README documents `MIRROR=ghcr.nju.edu.cn ./start.sh` to pull through a mirror and retag the official digest — verify with `docker image inspect` against the pinned digest in `compose.yaml`.\n\nDevelopers hacking on LoopForge itself use `bash scripts/dev.sh` with **Go 1.26.11+**, per the repository docs.\n\nSelf-hosted Docker installs keep sandboxes and approval history on your hardware\n\n## Executors, GitHub, and sandboxes\n\n**Executors** are the actual coding CLIs. CodeBuddy and Codex ship inside the image; Claude Code, OpenCode, Kimi Code, and Pi install on first use via npm and cache afterward. Cursor CLI and Antigravity use provider installers on chat first use but **are not preinstalled in SOP sandboxes**, the README warns — pick executors your sandbox can actually run.\n\n**GitHub** is optional: generic product SOPs can deliver documents without a repo; bug-fix and backend SOPs need source. OAuth device flow is built in; SSH keys work for GitLab, Gitea, or Gogs with manual PR tooling.\n\n**Sandboxes** talk to the host Docker socket — powerful and appropriate for **trusted teams**, not multi-tenant hostile users. That is the same trust model as [Chrome DevTools MCP](https://explainx.ai/blog/chrome-devtools-mcp-chrome-devtools-for-agents-55k-stars-setup-skills-guide-2026) driving a real browser: convenience trades off against isolation.\n\n## LoopForge Local for single developers\n\nTeams that want SOP ideas **inside** an existing CLI can install the portable edition:\n\n```\nnpx -y loopforge-cli@latest install claude\n```\n\nThen invoke `/start-devflow \"<your request>\"` in Claude Code. Source lives under `local/` in the main repository. Platform SOP export to standalone local runs is **planned**, not shipped — check the local guide before you depend on it.\n\n## How this compares to other orchestration stacks\n\n| Product | What it optimizes | \n|---|---|\n| [Dify / n8n / Flowise](https://explainx.ai/blog/what-is-a2a-agent2agent-protocol-mcp-comparison-guide-2026) | General AI automation graphs | \n| LoopForge | **Software delivery** with real repos, coding CLIs, PR evidence | \n| [Plannotator](https://explainx.ai/blog/plannotator-plan-code-review-workspace-coding-agents-claude-code-codex-github-trending-2026) | Rich human review of one plan or diff | \n| [Cognition Devin Apps](https://explainx.ai/blog/cognition-devin-apps-private-ai-built-web-tools-november-2026) | Hosted private micro-apps from agents | \n\nLoopForge is closer to **CI plus agent roles** than to a chat UI skin. If you already standardized on [Addy Osmani’s agent-skills packs](https://explainx.ai/blog/addy-osmani-agent-skills-production-pack-github-trending-2026), LoopForge gives those skills **stage boundaries** and **approval cards** in a shared inbox.\n\n## Token economics and observability\n\nThe README advertises **token and cost accounting** for SOPs and conversations at public list prices, plus engineering metrics exports. That matters when [coding-agent bills](https://explainx.ai/blog/track-and-cap-amazon-bedrock-spend-per-iam-principal-cur-2-0-aws-budgets-deny-policy-guide-2026) surprise finance — LoopForge makes spend visible per requirement, not only per API key.\n\nRuns expose **node status, tool traces, worker logs, and per-node artifacts**. When an agent step fails, you can **retry** or **rerun from any node** instead of replaying an entire chat from scratch — the operational difference between a toy demo and something a release manager will trust.\n\n## Security and governance notes\n\nReport vulnerabilities through Tencent’s published security policy on the repository. For production:\n\n- Treat Docker socket access as **root-equivalent** on the host.\n- Restrict signup with `AGENTFLOW_ALLOW_SIGNUP=false` when you do not want open registration.\n- Put HTTPS in front before exposing beyond a VPN.\n\nLoopForge adds **process** around agents that already pass your CLI policies; it does not replace dangerous-tool guardrails inside each executor. Pair organizational SOPs with tool restrictions the way [shadcn’s agent skill audit guidance](https://explainx.ai/blog/shadcn-improve-agent-skill-audit-expensive-execute-cheap-2026) recommends: expensive execute, cheap review.\n\n## HTTP API, MCP, and scheduled agent tasks\n\nBeyond the browser console, LoopForge documents an **OpenAPI-described HTTP API** and **MCP exposure** so external systems can open requirements, enqueue SOP runs, or poll gate status. That matters when your incident bot already lives in Slack: the approval card can be a webhook into LoopForge’s **Needs review** inbox instead of a second chat thread nobody audits.\n\n**AI scheduled tasks** reuse the same workspace machinery: recurring agent jobs with parameterized templates, versioned like SOPs, with full MCP and repository access when credentials allow. For teams comparing [Cursor’s proactive Projects coordinator](https://explainx.ai/blog/claude-code-projects-proactive-coordinator-november-2026), LoopForge is heavier ops but offers **shared** history — every teammate sees the same run graph.\n\n## Built-in SOPs and the validation record\n\nTencent ships **read-only blueprint SOPs** for backend, frontend, bug-fix, and product flows. Copy one into your project before you edit nodes — the README warns that built-ins are templates, not drop-in production paths. The repository links a **validation record** and **acceptance matrix** describing a real four-role CodeBuddy run (plan, implement, review, summarize) with three passing tests and a human gate — useful as a realism check, not a guarantee your repository will pass unchanged.\n\nIf you maintain internal “definition of done” docs, map each DoD bullet to a **Script node** (lint, unit tests, SBOM) so agents cannot skip checks by “forgetting” a prompt line.\n\n## Cost reporting and model selection per role\n\nLoopForge’s insights pages aggregate **token usage at public list prices** per requirement and per SOP version. That is not the same as invoice truth — your enterprise discount may differ — but it beats flying blind when [Gen Z stacks AI subscriptions](https://explainx.ai/blog/gen-z-ai-subscriptions-utility-bill-bank-of-america-2026) beside corporate API keys.\n\nPer-role executor choice is deliberate: a planner might run an expensive model while the implementer uses Codex on a smaller tier. The README notes that **requirement clarification, notification summaries, and scheduled assistant tasks** currently pin to **CodeBuddy-only model pickers** on the server — mixing CLIs on those subsystems is rejected rather than silently swapped. Read that constraint before you standardize on Claude Code everywhere.\n\n## Ops checklist before you trust production delivery\n\n| Step | Why | \n|---|---|\n| Pin compose digest | Mirrors can retag; verify `RepoDigests` after pull | \n| Restrict port 8090 | Default binds all interfaces | \n| Disable open signup | `AGENTFLOW_ALLOW_SIGNUP=false` for private teams | \n| Scope GitHub OAuth | Limit repositories per project | \n| Snapshot volumes | Re-running installer keeps data, but backups are still yours | \n| Test ZCode/Cursor paths | Not all CLIs exist inside SOP sandboxes | \n\n## When LoopForge is worth the ops cost\n\n**Choose LoopForge** if multiple developers share coding agents, you need **auditable** delivery (gates, artifacts, rerun history), and you can host Docker on a trusted machine.\n\n**Skip it** if one developer already lives entirely inside Claude Code with personal git hooks — [Worktrunk](https://explainx.ai/blog/worktrunk-git-worktree-cli-parallel-ai-agents-claude-code-2026) or Plannotator may be lighter.\n\n**Watch the mirror docs** if you deploy in regions where GHCR pulls stall; digest pinning is only meaningful if the pull actually completes.\n\n## Related reading\n\nSources: [Tencent LoopForge repository](https://github.com/Tencent/LoopForge?ref=explainx), [LoopForge product site](https://loopforge.cn/?ref=explainx)", "url": "https://wpnews.pro/news/tencent-loopforge-self-hosted-sops-that-orchestrate-claude-code-codex-and", "canonical_source": "https://explainx.ai/blog/tencent-loopforge-self-hosted-sop-coding-agents-claude-codex-2026?utm_source=rss&utm_medium=feed&utm_campaign=blog", "published_at": "2026-11-14 00:00:00+00:00", "updated_at": "2026-11-14 14:53:11.252714+00:00", "lang": "en", "topics": ["ai-agents", "developer-tools", "ai-tools", "artificial-intelligence"], "entities": ["Tencent", "LoopForge", "Claude Code", "Codex", "CodeBuddy", "Cursor", "GitHub", "Docker"], "also_reported_by": [], "alternates": {"html": "https://wpnews.pro/news/tencent-loopforge-self-hosted-sops-that-orchestrate-claude-code-codex-and", "markdown": "https://wpnews.pro/news/tencent-loopforge-self-hosted-sops-that-orchestrate-claude-code-codex-and.md", "text": "https://wpnews.pro/news/tencent-loopforge-self-hosted-sops-that-orchestrate-claude-code-codex-and.txt", "jsonld": "https://wpnews.pro/news/tencent-loopforge-self-hosted-sops-that-orchestrate-claude-code-codex-and.jsonld"}}