Is llm-space open source?

Yes — deer-flow/llm-space is open source, released under the MIT license.

What language is llm-space written in?

deer-flow/llm-space is primarily written in TypeScript.

How popular is llm-space?

deer-flow/llm-space has 875 stars on GitHub.

Where can I find llm-space?

deer-flow/llm-space is on GitHub at https://github.com/deer-flow/llm-space.

← all repositories

deer-flow/llm-space

macOS agent lab that replays your bot’s bad decisions

LLM Space exists because debugging an agent loop in a browser tab is tedious, and replaying a local trace is faster.

★875 stars TypeScript Coding Assistants LLMOps · Eval

View on GitHub ↗ Homepage ↗

Collecting fresh signals — velocity needs a few days of history.

collecting data…

star history

What it does

LLM Space is a macOS desktop app for prototyping and debugging LLM agents. You write prompts, configure tools, and run agent loops while the app traces every model call and tool invocation in real time. When a run fails, you replay the sequence from history and step through the harness to find the fault. All threads, files, and API keys stay in a local directory on your machine.

The interesting bit

The app ships in two editions: a lean build that piggybacks on the system WebView at roughly 27 MB, and a heavier “Performance” build that embeds its own renderer for consistency across macOS versions. Both read from the same ~/.llm-space data, so switching between them is painless. It is also the internal dogfooding companion for DeerFlow, ByteDance’s agent framework, which means the debugging workflow was forged against real agent failures.

Key highlights

Local-first by design: threads are stored as local files and API keys never leave your computer.
Replay debugging: pull any past run from history and step through the agent loop to isolate what broke.
Two macOS editions: a lightweight build using the native WebView, and a larger build with an embedded renderer for better performance.
Developed alongside DeerFlow, so the tracing and inspection features reflect actual agent-building needs.
Anonymous telemetry is collected by default, but the project documents exactly what is gathered and how to opt out.

Caveats

macOS only: releases are provided as DMGs for Apple Silicon and Intel, with no Windows or Linux builds mentioned.
The README advertises evaluation features but does not specify what metrics or scoring logic is used.
For now, the maintainers only merge pull requests from the DeerFlow core team; everyone else is asked to open issues.

Verdict

Grab it if you want a private, file-based sandbox for iterating on agents and diagnosing failures without browser DevTools. Look elsewhere if you need cross-platform support or a team-wide cloud workspace.

Frequently asked

What is deer-flow/llm-space?: LLM Space exists because debugging an agent loop in a browser tab is tedious, and replaying a local trace is faster.
Is llm-space open source?: Yes — deer-flow/llm-space is open source, released under the MIT license.
What language is llm-space written in?: deer-flow/llm-space is primarily written in TypeScript.
How popular is llm-space?: deer-flow/llm-space has 875 stars on GitHub.
Where can I find llm-space?: deer-flow/llm-space is on GitHub at https://github.com/deer-flow/llm-space.