A full-stack multi-agent system for subagent coordination, approvals, guarded execution, and tool-integrated workflows.
Alan Agentic Orchestrator is a multi-agent platform with a React and Next.js frontend, a FastAPI backend, multi-provider LLM routing, PostgreSQL and Redis-backed state, and external tool integrations through Composio.
The frontend identity of Alan is strongly React and Next.js, even though the orchestration backend is Python-heavy. That full-stack split is a major part of how the product works and how it feels.
Alan is built around decomposing complex work into sub-tasks routed to specialized workers.
The backend is split across provider routing, PostgreSQL and Redis storage, orchestration logic, prompts, Cerbos security policies, and specialized executors.
The frontend provides a chat-first interface with API-key management, approvals, markdown rendering, settings, and subagent execution views.
Supports code execution, shell commands, file operations, local system execution, and external-tool calls.
Uses provider-aware routing and failover for multiple LLM backends.
Applies sandboxing and approval boundaries for more controlled local execution.
Builds a real-time chat experience around orchestrated AI workers rather than exposing orchestration as a hidden backend detail.
Lets users inspect approvals, settings, provider state, and subagent progress visually.
Makes external tool integrations part of the runtime rather than isolated demos.
Cerbos policies, Redis locks, and real-time UI components all contribute to the feeling of a serious orchestration environment.
Delegates complex work across code execution, command execution, file operations, local execution, and tool-integrated subagents.
Combines provider routing, PostgreSQL persistence, Redis caching and locking, orchestration prompts, and access policies.
Includes a chat interface with approvals, settings, markdown rendering, and subagent trace views.
A user submits a task through the frontend chat surface.
The orchestrator decomposes it into specialized subagent work and routes to the right executor or external tool.
State, messages, and steps are stored while the UI reflects execution and approvals in real time.
The product has a strong full-stack identity because the orchestration engine and the operator-facing chat surface are both treated as first-class.
Alan was approached as a full orchestration product rather than a chat interface with hidden complexity behind it. Task decomposition, worker routing, approvals, provider management, and execution safety all need to feel coherent together for a multi-agent system to be trustworthy. That is why the product puts equal weight on the operator-facing UI and the orchestration engine underneath it.
Within the broader project set, Alan represents the collaborative and operational side of agent systems. It extends the same interest in structured execution and observability, but it does so through a browser-based interface with multi-user implications, persistent state, and approval-aware flows. That gives it a distinct role as a more team-facing orchestration environment.