Who Orgtree is for#
Developers directing parallel AI coding agents
Orgtree's canvas, docket, and mail system give you a coordination layer for directing multiple Claude Code or Codex agents concurrently. Assign each agent a role and charter, delegate subtasks through the docket, and read concise progress reports in your inbox rather than scrolling through full conversations.
Skip if:
Skip if you are running a single agent on a single task. The overhead of setting up charters, credits, and a docket pays off at two or more parallel agents. For one-shot tasks, using Claude Code or Codex directly is simpler.
Small teams sharing AI agent infrastructure
A standalone mail hub lets a small team connect their individual Orgtree organizations so agents on one machine can message agents on another. Pair with the Hubchat companion app to monitor and message agents from a phone over your private network.
Skip if:
Skip if the team needs real-time collaborative editing or shared live context rather than asynchronous delegation. Orgtree is built around inbox-based coordination between agents, not simultaneous access to a shared workspace.
Developers mixing cloud and local models
The language server provider lets you connect Ollama, LM Studio, or vLLM agents alongside Claude Code or Codex agents in the same organization. Assign GPU-local models to cost-sensitive bulk tasks and cloud models to tasks that need higher capability, with per-agent account selection.
Skip if:
Skip if your local hardware cannot support at least 65,536-token context windows. Orgtree flags smaller context windows as insufficient for its agent instructions and tool definitions, and models running at 4K context will produce degraded results.
Power users running persistent long-running agent teams
Orgtree stores organizations and agent conversations in a local PostgreSQL database that survives individual sessions. Watchdog agents can wake on file changes or process events rather than polling, and system-tray operation keeps work moving between your visits over days or weeks.
Skip if:
Skip if your workflow is a series of one-off tasks with no need for context continuity between sessions. Orgtree's persistent storage and recovery machinery matter when the same agents need to resume work after restarts.
The problem it solves#
Running multiple AI coding agents in parallel has no natural coordination layer. Each agent lives in its own terminal or chat window, with no shared task record, no way to delegate subtasks between agents, and no place to track decisions and context as the work moves between sessions. Following several agents at once means scrolling through hours of tool calls across separate windows, and there is no unified way to send direction to an agent mid-task without interrupting its flow.
The challenge compounds over days: agents lose context when sessions restart, work items have no owner, and the person directing the team must hold the whole job state in their own head. What starts as a small efficiency gain from AI assistance becomes a coordination burden that grows with the number of parallel agents.
How it solves it#
Visual org-chart canvas
View the full agent team as a row layout or circular org chart on a drag-and-drop canvas. Rearrange agents between teams as the job changes. Open any agent's desk to read its live conversation and inspect tool calls without switching windows.
Persistent work docket with tickets
Each task in the docket has an owner, requirements, status, and a progress record. Work items stay with a task when it changes hands between agents, keeping decisions and attachments in one place across sessions and agent handoffs.
Agent mail and inboxes
Every agent has a persistent inbox. Agents send progress reports and questions directly to your inbox so you can follow work without scrolling through full conversations. Reply while agents continue working, or grant a direct audience outside the reporting chain. Organizations on separate machines connect through a shared mail hub.
Multi-provider and local model support
Run Claude Code, OpenAI Codex, Google Antigravity, Muse Code, and OpenRouter agents in the same organization. The language server provider connects Ollama, LM Studio, llama.cpp, and vLLM so you can run locally hosted models alongside cloud providers in the same agent team.
Credit-based capacity budgets
Each model has a seat cost, and each agent has a credit grant it can use to staff a team beneath it. This controls how many agents can be active at once and how deep the delegation chain can go, independently of provider billing and token limits. Retiring agents releases that capacity.
Context compaction for cold turns
When a session's context exceeds a configured threshold before a known-cold turn, Orgtree starts a fresh session seeded with a summary. This avoids re-reading long conversation histories when the prompt cache has expired, reducing inference cost on long-running agents.
Strengths and trade-offs#
Strengths
- MIT license with local-only data storageThe MIT license permits commercial use, modification, and redistribution without restriction. Organizations, conversations, mail, and dockets are stored in a bundled PostgreSQL database on your own machine. Unlike SaaS orchestration services with per-seat fees, the software itself has no ongoing license cost.
- Coordination depth beyond single-agent CLIsAgents can hire sub-agents up to their credit grant, exchange typed mail, share a work docket, and pass questions up the hierarchy. A Needs Attention view consolidates open requests, urgent mail, and flagged tickets in one place. This coordination layer does not exist in standard CLI agent scripts or simple chat interfaces.
- Background operation and session recoveryOrgtree can start at Windows login and keep its engine running in the system tray after you close windows. On restart, it recovers pending mail and sends continuation messages to interrupted agents, so work continues between your sessions without manual re-queuing.
- Local model hosting via language serversOllama, LM Studio, llama.cpp, and vLLM connect through the language server provider. Claude Code or Codex runs on the Orgtree machine as the agent CLI, while inference comes from the configured server. GPU-local capacity can substitute for cloud API cost on cost-sensitive bulk tasks.
Trade-offs
- -Windows-only installerThe published installer runs on 64-bit Windows. There is no documented macOS or Linux build in the project's GitHub releases. macOS and Linux users cannot run the desktop application.
- -Steep learning curveThe README describes Orgtree as 'a substantial tool, with a learning curve to match.' Effective use requires understanding canvas controls, charter setup, credit budgets, and docket workflow before the organizational model pays off. The project itself recommends starting with one agent and a small task.
- -Early-stage project with limited historyOrgtree was created in September 2026 and has 146 GitHub stars. The feature set is growing quickly, and users should expect rough edges and possible breaking changes between releases at this early stage of the project.
Orgtree vs alternatives#
Orgtree vs Microsoft Copilot Studio
Both tools address multi-agent AI orchestration, but they serve different deployment models. Microsoft Copilot Studio is a cloud-hosted, enterprise-grade service built on Azure; Orgtree is a Windows desktop application that stores everything locally.
| Feature | Orgtree | Microsoft Copilot Studio |
|---|---|---|
| License | MIT | Proprietary |
| Deployment | Local Windows desktop | Cloud (Microsoft Azure) |
| Agent state storage | Local PostgreSQL | Microsoft cloud |
| Pricing | Free software | Per-message and per-session billing |
| Model support | Claude Code, Codex, OpenRouter, local | Microsoft-approved models |
Orgtree is the better choice when data privacy, model flexibility, and zero orchestration-layer cost matter. All agent state and conversation history stays on your machine, and you can mix cloud providers with locally hosted models in the same organization. Copilot Studio is worth considering when you need enterprise governance, Azure Active Directory integration, pre-built Microsoft 365 connectors, or managed cloud execution that Orgtree does not provide.
Orgtree vs CrewAI Cloud
CrewAI Cloud offers a managed SaaS environment for multi-agent workflows built with the Python CrewAI framework. Orgtree covers the same coordination concept as a self-hosted Windows desktop application.
| Feature | Orgtree | CrewAI Cloud |
|---|---|---|
| License | MIT | Proprietary SaaS |
| Deployment | Local Windows desktop | Cloud-hosted |
| Model providers | Claude Code, Codex, OpenRouter, local | API-based (various) |
| Pricing | Free software | Paid SaaS tiers |
| Storage | Local PostgreSQL | Vendor-managed |
Orgtree is preferable when you want to use Claude Code or Codex as the agent CLI with their native tool-calling capabilities, keep all work data on your own machine, or avoid a SaaS subscription for the orchestration layer. CrewAI Cloud suits teams already building with the Python CrewAI framework or those who need managed hosting without running a desktop application.
What it's built on#
- Languages
- JavaScriptPythonRustTypeScript
- Frameworks
- React
- Tooling
- esbuild
FAQ#
Is Orgtree free to use?
Yes. Orgtree is MIT licensed and the installer is free to download from GitHub. There is no license fee for the orchestration software itself. You still pay for provider subscriptions (Claude Code, Codex, OpenRouter API keys) that you connect to it, which have their own billing.
Does Orgtree work on macOS or Linux?
No. The current installer targets 64-bit Windows only. There is no documented macOS or Linux build in the project's GitHub releases as of October 2026.
What AI providers does Orgtree support?
Orgtree supports Claude Code (Anthropic), Codex CLI (OpenAI), Google Antigravity, Meta Muse Code, and OpenRouter. It also supports local models via Ollama, LM Studio, llama.cpp, and vLLM through its language server provider. You configure only the providers you plan to use in App settings.
How does context compaction work in Orgtree?
When an agent's context exceeds a configured threshold before a known-cold turn, Orgtree starts a fresh session seeded with a summary of the prior conversation. This reduces inference cost by avoiding a full re-read of a long conversation when the prompt cache has expired. You enable it per agent in its settings.
Can I connect Orgtree agents across machines for a team?
Yes. A standalone mail hub can run on a shared server accessible to the whole team. Each developer connects their local Orgtree organization and the Hubchat app to the hub, letting agents on different machines exchange messages and files asynchronously. The built-in hub in the desktop app is designed for individual or small team use.
Similar open-source tools#
Hypa
Local-first terminal workspace for AI coding agents
Hermes Agent
Self-hosted AI agent with persistent memory, multi-channel chat, and model choice across OpenAI, OpenRouter, and custom endpoints.
Agent Cache Web
Convert docs URLs into Markdown bundles for AI coding agents
iFixAi
Self-hosted AI agent auditing across 60 trust and safety inspections.
WeKnora
Self-hosted knowledge base with RAG search, AI agent, and wiki
Marker
Convert documents to clean markdown and JSON with open source AI.