{
  "id": "page:docs-reference-repos-sero-labs-sero-research",
  "_kind": "Page",
  "_file": "wiki/docs/reference-repos/sero-labs/sero/research.md",
  "_cluster": "wiki",
  "attributes": {
    "nodeKind": "Page",
    "sourcePath": "docs/reference-repos/sero-labs/sero/research.md",
    "sourceKind": "repo-docs",
    "title": "sero-labs/sero",
    "displayName": "sero-labs/sero",
    "slug": "docs/reference-repos/sero-labs/sero/research",
    "articlePath": "wiki/docs/reference-repos/sero-labs/sero/research.md",
    "article": "\n# sero-labs/sero\n\n## Metadata\n- **Stars:** ~2,000\n- **License:** AGPL-3.0\n- **Last pushed:** 2026-06 (actively maintained)\n- **Topics:** agent-os, desktop, electron, pi-framework\n- **Fork:** No\n- **Author:** Sero Labs (Mario Zechner / badlogic)\n\n## Archetype: agent-workspace-platform\n\nA personal agent operating system (Agent OS) built on the Pi coding agent framework. Electron desktop app that consolidates chat, terminals, file explorer, visual browser, and self-extending plugins into a unified local-first workspace where agents maintain persistent awareness across sessions.\n\n## Structure\n- `apps/desktop/` — Electron + React shell (main process + renderer)\n- `packages/` — Shared runtime, UI, and common libraries\n- `plugins/` — Built-in and example plugins (scout, reviewer, test-writer)\n- `eval/` — Promptfoo-based evaluation framework\n\n## Tech Stack\n- TypeScript (96%), Electron, React, Module Federation\n- Built on Pi coding agent framework (@earendil-works/pi-*)\n- Node.js 22, pnpm 10\n- Platforms: macOS (Apple Silicon), Linux (x64/arm64), Windows (x64)\n\n## Key Concepts\n\n### Agent-as-Workspace\nThe fundamental inversion: instead of an agent living in a sidebar/tab, the agent IS the workspace. Files, terminals, browser, and plugins all exist as native context the agent can see and reason about without manual context injection.\n\n### Self-Extending Plugin Paradigm\nThe agent builds plugins on demand during conversation. User describes a workflow → agent scaffolds plugin (manifest + tools + optional React UI panel) → plugin is live immediately via Module Federation, no restart. Plugins persist across sessions and can be iteratively refined.\n\n### Content vs Details Tool Response Pattern\nTool results separate LLM context from UI rendering:\n- `content`: Clean text for the model to reason about\n- `details`: Structured data (JSON, images, charts) for the UI to render\nNeither pollutes the other. Solves the \"parsing text for UI\" problem at the source.\n\n### Session Tree (Not Linear Chat)\nConversation history is a tree with fork/clone/branch semantics. Each message has a parentId. Users can explore alternative approaches without losing the original path. Context is linearized root-to-leaf for LLM APIs. Dead-end branches can be compacted independently.\n\n### Composable Hook Interceptors\nExtensions intercept the agent loop at surgical points (beforeToolCall, afterToolCall, message_start, turn_end). Enables: plan mode (approve before executing), git checkpointing (auto-commit before mutations), custom compaction triggers, safety filters. Hooks compose — multiple extensions can register for the same event.\n\n### Markdown Specialist Agents\nBuilt-in agents (scout, reviewer, test-writer) are plain markdown files with YAML frontmatter (name, tools, model, permissions) and markdown body (system prompt). Editable, forkable, composable. The agent can create new specialist files during conversation.\n\n### Git-Checkpointed Compaction\nBefore summarizing old conversation turns, create a git commit. The agent can reconstruct state via `git log` and `git diff` after compaction. Compaction becomes reversible — information moves from conversation context to git history.\n\n### Visual Reasoning Loop\nBuilt-in browser enables closed-loop visual feedback: agent writes code → code renders in browser → agent takes annotated screenshot (elements get numbered labels) → agent reasons about visual layout → agent fixes issues → repeat.\n\n## Extraction Priority: High\n\n### Extractable Patterns\n\n| Pattern | Priority | Notes |\n|---------|----------|-------|\n| Self-extending plugin paradigm | High | `createPlugin` agentic tool + Module Federation hot reload |\n| Content/details tool response | High | Clean separation of LLM context from UI rendering |\n| Session tree with fork/clone | High | Non-linear conversation exploration |\n| Composable hook interceptors | Medium | Plugin-registered beforeToolCall handlers |\n| Visual reasoning loop | Medium | Annotated screenshots + vision-based feedback |\n| Git-checkpointed compaction | Medium | Auto-commit before context compression |\n| Markdown agent creation | Medium | Agent creates specialist agents mid-conversation |\n| Mosaic workspace layout | High | Split-pane desktop (chat + terminal + files + browser) |\n\n### What We Already Have (No Extraction Needed)\n\n| Sero Feature | Our Equivalent |\n|---|---|\n| Multi-provider LLM | adapters/core — 25+ providers, transport-adapter proxy |\n| Plugin system | genty/platform/plugins — sandbox, permissions, marketplace |\n| Memory/persistence | sdk/profiles + sdk/session — user+project profiles, session state |\n| Subagent framework | genty/core/subagent — 3 invocation modes, oversight |\n| MCP integration | genty/platform/mcp — full client with channels, permissions |\n| Container isolation | Docker, SSH, K8s transports, secure-sandbox |\n| Agent definitions | .claude/agents/ + AgentPersona/Soul/Definition CRDs |\n| Multi-platform UI | genty/{desktop,web,tui,mobile,tv,watch} — 7+ surfaces |\n\n### What We Have That Sero Doesn't\n\n- Kubernetes-native forge (Kradle) with 89+ CRD types\n- Process definitions + babysitter orchestration with breakpoints\n- Multi-agent Jitsi meeting integration\n- Agent identity system (Persona, Soul, Appearance, Voice profiles)\n- Atlas knowledge graph\n- 7 platform surfaces (vs Sero's 1)\n- CI/CD integration (GitHub Actions workflows)\n- Governance layer (authority chains, mandates, posture-based permissions)\n- Agent marketplace / process library\n\n## Related Files\n- [sero-gap-analysis.md](./sero-gap-analysis.md) — Feature-by-feature gap analysis\n- [sero-concepts-and-opportunities.md](./sero-concepts-and-opportunities.md) — Deep architecture concepts and build opportunities\n",
    "documents": []
  },
  "outgoingEdges": [],
  "incomingEdges": [
    {
      "from": "page:docs-reference-repos",
      "to": "page:docs-reference-repos-sero-labs-sero-research",
      "kind": "contains_page"
    }
  ]
}