Describe what you want.
Ask for a feature, bring a bug, or explore an idea. Your agent can read the project files and help you figure out where to start.
AI DEVELOPMENT WORKSPACE · OPEN BETA v0.1.13
Build something incredible.
Minnow brings the tools you need to create into one workspace: AI agents, Code, Source Control, Issues, Brain, and more. All in one desktop app you can extend and customize.
Connect Claude Code, Codex, or Cursor, use a cloud API, or run a model on your own computer.
Free and open source. No Minnow account. Local or cloud models.
INSIDE MINNOW Notes Sync: a failing test, a one-file fix, and the diff ready to review.
ONE WORKSPACE, BY DESIGN
Built to work together.
Your agent can read an issue, look up a decision in Brain, work on the code, and run your tests. You use those same tools to follow along, make changes yourself, and review the result.
See Minnow at workCode
Write and edit your software.
Source Control
Review and manage your changes.
Issues
Track bugs, ideas, and work to do.
Brain
Keep notes and decisions.
One shared workspace
Code, Source Control, Issues, and Brain all work with the project you have open.
THE BUILD LOOP
See Minnow at work.
Start with a project and something you want to do. Work directly with an agent, write code yourself, or use a board to divide a larger job between agents.
Work out a plan.
Talk through the approach, ask questions, and adjust the scope. Save the plan when you’re ready to build.
Build it.
Work alongside an agent in Code. For larger jobs, use a board to assign building and testing tasks to multiple agents. With worktree isolation enabled, each task gets a separate working copy of the repository.
Commands run on your machine. Enable the agent sandbox for unattended runs.
Review the changes.
Read the diff and check the test results in Source Control. Make another pass, commit the work, or open a pull request.
Keep what you learned.
Save useful decisions in Brain and follow-up work in Issues. Your agents can refer to those notes and issues when you return to the project.
THE APPS
Get to know your workspace.
Start in Code. Review in Source Control. Keep decisions in Brain. Explore the tools your agent can use without leaving the project.
01 · Code
The editor and the agent work on the same repo.
File tree, LSP, inline completion, Quick Edit, terminal, dev servers, and a Chromium preview.
Chat works directly on the repo. Use Quick Edit to preview a suggested change, or review agent edits in Source Control.
- LSP
- inline completion
- Quick Edit
- terminal
- dev servers
02 · Source Control
Git without leaving the workspace.
Review changes, history, branches, stashes, worktrees, pull requests, and CI.
The agent can commit and open PRs. You still decide what ships.
- changes
- branches
- worktrees
- PRs + CI
03 · Orchestrator
Turn a plan into parallel agent work.
A plan becomes waves of Builder and Tester tasks.
With isolation enabled, each task runs in its own worktree. Completed tasks merge into an integration branch; review the assembled result before committing it to your destination branch.
- plan → waves
- Builder + Tester
- worktrees
- integration branch
04 · Models
Run local models, cloud models, or both.
Hardware-fit scoring tells you whether a model will run well on your machine before you download it.
Serve models with llama.cpp or MLX, connect a cloud API, or use Claude Code, Codex, or Cursor CLI. Assign models by role, or use a router to share capacity and fail over.
- hardware fit
- llama.cpp + MLX
- CLI providers
- model routers
05 · Brain
Your project gets a memory.
Brain is a Markdown knowledge base with semantic search, a code index, and memories the agent reads and writes.
Every page is a file on your disk. What one session works out, the next one starts with.
- Markdown wiki
- semantic recall
- code index
- memories
06 · Issues
The agent files what it finds.
List, board, triage, and saved views, plus issue tools the agent can use.
If an overnight run hits an edge case, the issue is waiting for you in the morning with steps to reproduce.
- list + board
- triage
- saved views
- agent-filed
07 · Scheduler
Put repetitive work on a schedule.
Run agent jobs on an interval or a cron schedule, scoped to a workspace, with a history for every run.
Jobs run while Minnow is running, including in the system tray. Quitting the app stops scheduled execution.
- intervals + cron
- run history
- workspace scope
- /loop
- /goal
Bring your model.
Local runtimes, cloud APIs, and agent CLIs. Choose what runs your agents.
No Minnow account. No Minnow subscription. Minnow puts no cap on usage. Cloud providers set their own prices and limits.
FEATURES
Built in, across the whole workspace.
Every surface shares the same agent, tools, and models, so these work the same wherever you are in the app.
-
Modes
General, Build, Plan, and Debug. Each mode sets the prompt and available tools. Plan mode restricts file-edit tools to plan documents; shell commands still follow your permissions.
Learn more → -
Sub-agents
Hand a job to a sub-agent: explore the codebase, run shell commands, debug a failure, or write up an issue. It works in the background and reports back, so your conversation stays short.
Learn more → -
Permissions
Every built-in tool can be set to Full, Ask, or Off. Each call shows up in the transcript as it happens.
Learn more → -
Skills
Reusable workflows like code review, opening a PR, and writing tests. Type / to run one. Each is a Markdown file you can edit or copy.
Learn more → -
MCP and plugins
Connect any MCP server or local plugin. Their tools follow the same permission rules as the built-in ones.
Learn more → -
Code intelligence
The agent can map the repo, find symbols and their callers, and read LSP diagnostics instead of searching text blindly.
Learn more →
COMPLETELY YOURS
Your data is just files.
Chats, settings, memories, and downloads live under ~/.minnow, as Markdown, JSON, and SQLite.
Nothing is locked in a format only Minnow can read.
-
Edit anything by hand
Skills, permissions, rules, and prompts are plain text. Edit them in any editor and the next session picks up the change.
-
See what the model sees
Mode prompts and Quick Edit prompts are files too, so you can read exactly what gets sent to the model and change it.
-
No telemetry. No Minnow account.
Minnow collects no telemetry and needs no Minnow account. API keys are encrypted on your disk. Cloud models receive the prompts and context sent to them.
-
Back it up and take it with you
Create a backup in Settings, or schedule daily or weekly snapshots. A passphrase-protected backup can include your credentials. Restore it on another computer and keep your project knowledge.
~/.minnow
├── .key
├── config.json
├── tools.json
├── sessions/
├── brain/
├── issues/
├── prompts/
├── skills/
├── providers/
└── models/
ABOUT MINNOW
Why I’m building Minnow
Minnow started as a simple chat app for Local AI. I kept adding the tools I use and trying ideas, what resulted is Minnow today: a workspace with everything I needed to build in one place.
I want to make something in the spirit of Blender. A tool anyone can use, without paying for access to the parts they need. Something you can pick up, make your own, and use to build what you have in mind.
I’m building it for myself, and sharing it so other people can use it too.
Supported by the people who use it
Minnow is licensed under AGPL-3.0-or-later. There's no Pro tier, no per-seat pricing, and no feature that unlocks when you pay.
Sponsoring supports the work. It doesn't unlock anything extra.
FAQ
Things worth knowing before you download it.
What is Minnow?
Minnow is a desktop development workspace with an agent built in. The editor, git, issues, planning, Brain, scheduling, and model hosting share one context, so the agent can work across all of them. It's free, open source, and built for one person on one machine.
Do I need an API key?
No. Run a local model through llama.cpp, MLX, LM Studio, or Ollama. You can also connect an installed Claude Code, Codex, or Cursor CLI using its supported sign-in. Cloud APIs use your own key. Vendor prices, subscriptions, and usage limits still apply; Minnow adds no subscription or markup.
What hardware do I need?
Minnow runs on 64-bit Windows 10 or later, macOS 12 or later on Apple Silicon, and 64-bit x86 Linux. Cloud APIs and agent CLIs do not need local model hardware. For local models, Models estimates memory use from your CPU, RAM, GPU, and selected context before you download. Actual usage depends on the model and runtime settings.
Does my code leave my machine?
Only where you send it. With a local model, prompts and code stay on your machine. With a cloud provider, that provider receives your prompt and the context you include. Minnow also connects out when you ask it to: web searches and page fetches, Hugging Face downloads, GitHub through your gh CLI, MCP servers you add, webhooks you configure, and update checks against GitHub Releases. There's no telemetry.
How is Minnow different from an AI code editor?
Editors like Cursor, or VS Code with Copilot, put an agent in the editor. In Minnow, the editor is one part of a larger workspace, so the agent working in a file can also read the related issue, check the branch, look up past decisions in Brain, and file what it finds.
Can my team use it?
Everyone on a team can install their own copy, and the license allows it. Minnow is built for one person on one machine, though: there's no shared workspace, admin console, or SSO. Where a feature could be built for solo use or for teams, the solo version came first.
What does AGPL mean for me?
You can run it, read the source, change it, and fork it. If you distribute a modified version, or run one as a network service for others, you have to share that source under the same license. If you're just using Minnow on your own code, the license doesn't ask anything of you.
Start building.
Available for Windows, macOS, and Linux. No Minnow account needed, and no API key if you use a local model. Open a project, connect your model, and make your first change.