v1.2.0 · Open source · Free

Build your AI team.You direct the CEO.

You set the goal and direct the CEO agent. The CEO orchestrates the rest of the team. We’re preparing hierarchical multi-agent integration so you can build your team in ConductLoop.

macOS 13+ · Apple Silicon · Signed & notarized · Free under MIT

YOUR MULTI-AGENT TEAM · IN DEVELOPMENT
ConductLoop / Your AI teamIN DEVELOPMENT
HIERARCHICAL ORCHESTRATION

You lead. The CEO orchestrates.

Build your team of agents around a goal.

YOU

Set the goal and direct the CEO

CEO agent

Orchestrates the work of the entire team

CEO
Agent ACoordinated by the CEO
Agent BCoordinated by the CEO
Agent CCoordinated by the CEO
A shared context foundation for the team

Your notes, repositories and sources give context to your agents’ work.

Concept diagram · hierarchical integration in development
Runs on 127.0.0.1Bring your own modelMIT licensed ↗

A HOME FOR THE TOOLS YOU ALREADY USE

◈ ObsidianSQLite⌘ MCPOllamaClaudeOpenAI

What's inside

The foundation for your AI work.

Today, work with Brain, organize agents and tasks in Boardroom, and connect tools through MCP. We’re preparing the next step: a hierarchical team where you direct the CEO and the CEO orchestrates the agents.

01

Boardroom

Set up project agents, tasks and runs around the material in your local index. Review pending work and scheduled activity from one cockpit.

02

Brain chat, BYOK

Ask questions about indexed work and inspect the cited sources. Use a key for Anthropic, OpenAI or MiniMax, or point it at a local Ollama model.

03

MCP server

Expose search, recall and capture tools to MCP clients such as Claude Desktop or Cursor. They can use the same local memory as the app.

04

Context for your agents

Index an Obsidian vault, notes, repositories, skills, playbooks and URLs in local SQLite. Search with full-text and fuzzy matching, with optional local semantic embeddings.

05

Obsidian, both ways

Sync memory with plain Markdown in your vault. Keep your notes readable and editable with the tools you already use.

06

Capture & briefings

Capture ideas into an inbox, import an ICS calendar and prepare a briefing from your local project context.

Start with your work

Bring the context
you already own.

ConductLoop indexes the files and sources you choose. It gives your projects a shared local memory without moving your working files into another hosted workspace.

SQLite·127.0.0.1
01

Notes and vaults

Index an Obsidian vault and local memory folders, then search across the material you have collected.

02

Projects and code

Bring repositories, skills and project playbooks into the same local index as your notes.

03

Things worth following

Add URLs to watch and calendar files to import, alongside the work already on your Mac.

04

A memory your tools can reach

Use Brain chat and project agents in ConductLoop, or expose search and capture through its MCP server.

How it works

One local index.
A workflow you can inspect.

01

Choose what to index

Add the Obsidian vault, repositories, memory folders and URLs you want ConductLoop to search.

The local server binds to 127.0.0.1 by default. Your original files stay where they are.

02

Pick a model when you need one

Use your own Anthropic, OpenAI or MiniMax key, or connect Ollama on your Mac.

ConductLoop does not proxy model calls. Prompts and selected context go to the provider you configure.

03

Use and review the context

Ask the Brain questions with citations, organize project work in the Boardroom, or let another MCP client search and capture.

The index and app data live in local SQLite; Obsidian sync uses plain Markdown.

Know where things run

Your memory stays
on your machine.

ConductLoop is local-first. Its data flow is simple to inspect, and the connections it makes depend on the providers and tools you configure.

⌂On your Mac by default
✓Your indexed content and app data
✓The SQLite index and memory
✓Your original notes and repositories
↗Only when you choose
↗Prompts and context sent to your chosen model
↗Connections you enable for other services
↗Optional error reporting, only if configured

Open source

Built in the open.
Audit it, fork it, improve it.

The whole desktop app — Python engine, web UI and macOS shell — is public. If you want to know what it does with your notes, read the code. If it's missing something, send a pull request.

MIT

A real open-source license

Not source-available, not a community edition. Use it, change it and redistribute it under the license in the repo.

0

Accounts required

No sign-up, no email gate, no license key. Telemetry is off — error reporting only runs if you set it up yourself.

127.0.0.1

Where it runs

The server binds to your own machine by default. Your vault, keys and database stay there unless you send them to the model you picked.

Guides for your team

From idea to assignment.

Explore all guides →

ConductLoop

Frequently asked questions

What is ConductLoop?+
ConductLoop is a free, open-source AI workspace for macOS. It brings project agents, tasks, models and local context together through Brain, Boardroom and MCP. Hierarchical multi-agent integration is in preparation.
How will the CEO agent and multi-agent team work?+
You will build your team and direct the CEO agent. The CEO will orchestrate the work of the other agents around your goal. This hierarchical integration is in preparation and is not included in the current v1.2.0 download; no release date is announced here.
What can I use today?+
The current app includes Brain chat with cited sources, Boardroom project agents and tasks, an MCP server, local search, Obsidian sync, capture and briefings. The planned CEO-led team extends this foundation.
Is ConductLoop really open source?+
Yes. The desktop app — the Python engine, browser UI and macOS shell — is public on GitHub under the MIT license. You can read it, build it yourself, fork it and redistribute it under that license.
Is it free? What's the catch?+
It's free: no trial, subscription, account or email gate. You pay your model provider directly if you use a cloud model (Anthropic, OpenAI or MiniMax), or nothing for model usage with local Ollama. ConductLoop itself does not proxy model calls.
Where does my data go?+
The index and app data stay in local SQLite. Prompts and selected context go to the model provider you configure; with local Ollama, model calls stay on your Mac. Configured integrations may contact external services. Error reporting only runs if configured.
Explore all frequently asked questions →

Free. Open source. Yours.

Download the signed macOS app or build it from source. No account, no email, no payment.

MIT licensed · No account, subscription or email gate.

Release notes

The app updates itself. To hear about each new version anyway, watch the releases on GitHub or add the RSS feed to your reader.

Chat on WhatsApp