llm-space
A desktop app to prototype agent ideas, inspect every harness step, replay failures, and evaluate performance, all in one place. Local-first, cloud-ready for managed agents.
What is it?
What it is
LLM Space is a desktop application that lets users build, prototype, trace, debug, evaluate, and manage AI agents locally. It provides a unified UI for writing prompts, system messages, tools, and model settings, and it records every model call and tool run in the agent loop.
Why it exists
The project was created to give developers a single, local environment for experimenting with agent workflows, inspecting execution details, and iterating quickly. It also serves as a dog‑food tool for the DeerFlow project, ensuring every DeerFlow release is built and debugged with LLM Space.
Who should use it
Developers, AI engineers, and researchers building LLM-based agents who want a local desktop tool for prototyping, tracing, debugging, and evaluating agent workflows.
Who should avoid it
Non-technical users seeking no-code solutions, those who prefer cloud-hosted agent platforms, or anyone unwilling to install and maintain a desktop application.
How it works
A quick walkthrough in plain English
How llm-space works in 4 steps
Step 1 of 4
Something triggers the flow
A schedule, webhook, or manual click tells llm-space to start.
Features
Advantages
- All-in-one desktop app for prototyping, inspecting, debugging, and evaluating agents
- Keeps files and API keys locally on your machine
- Open source MIT license
- Dogfooded by DeerFlow core team ensuring real-world reliability
- Built with modern stack: TypeScript, Bun, Electrobun, React, Tailwind CSS, shadcn/ui, Pi Agent Core
- Two editions: lightweight (~27 MB) using system WebView and performance (~130 MB) with embedded renderer
- Shared data directory allows seamless switching between editions
- Self-updating applications
- Optional anonymous telemetry to improve app (opt‑out available)
Disadvantages
- Currently distributed as macOS DMG only (Apple Silicon and Intel)
- Building from source requires Bun installation
- External contributions limited to DeerFlow core team members (others can only open issues)
- Collects anonymous usage data (though opt‑out is provided)
- Performance edition has larger download size (~130 MB)
Installation
native
Grab a DMG from the latest release — macOS, Apple Silicon and Intel. It comes in two editions: LLM Space (uses the system WebView) or LLM Space Performance (embeds its own rendering engine).
FAQ
How do I download LLM Space?
Grab a DMG from the latest release page. There are two editions: LLM Space (uses system WebView, ~27 MB) and LLM Space Performance (bundles its own rendering engine, ~130 MB). Both share the same ~/.llm-space data folder.
What do I need to build LLM Space from source?
You need Bun installed. After installing Bun, run `bun install` from the repository root. For the exact CI toolchain, install mise and run `mise run setup`.
How do I start the desktop app for local development?
Run `mise run dev` (or `bun run dev` if you prefer) to start the desktop app in development mode.
Where are my threads, settings, and API keys stored?
All user data stays on your local machine in the `~/.llm-space` directory. LLM Space does not upload your files or API keys.
Can I contribute to LLM Space if I’m not on the DeerFlow core team?
Direct pull requests are currently limited to DeerFlow core team members, but you can help by opening issues for bug reports, feature ideas, and feedback.
Featured in Videos
YouTube tutorials and walkthroughs for llm-space
Alternatives
Similar projects ranked by category, topics, and text overlap.
