TL Studio
A local-first coding workspace built around a Native Agent. Edit, search, run, preview, chat, use tools and connect models from one workspace on your own machine.
src/
app.ts
styles.css
SEARCH
TERMINAL
tools: owned,
sessions: owned,
permissions: owned,
cloudBackend: none
};
Reading project…
✓ files
✓ tools
✓ permission
What changed
TL Studio is no longer “agent chat with some tools attached.” The backend owns the complete model → tool → model loop, sessions, permissions, project mutations and semantic events.
The current stable release is v0.6.0; active development is on v0.7.0-alpha.1.
What it does now
The product owns agent execution, cancellation, loop guards, persistence and permission-gated tools instead of delegating the core workflow to another coding runtime.
Monaco editing, project search, file operations, terminal processes and isolated preview live in the same local product boundary.
Native support covers OpenAI-compatible APIs, OpenAI Responses, Anthropic Messages and Gemini. Supported account-backed paths include ChatGPT/Codex and Claude alongside manual API configuration.
External tools join the same Tool Registry and permission boundary instead of becoming a separate automation island.
Stable builds target Windows x64, Linux x64/ARM64 and macOS Intel/Apple Silicon, with an npm launcher for quick project startup.
Boundary I care about
TL Studio does not require a project-owned hosted backend, database, telemetry service or model proxy. Project files stay local; model traffic goes directly to the provider the user configures.
The repository remains the source of truth for release status, security notes and architecture.
Stable v0.6.0 · Dev v0.7.0-alpha.1
Use the stable release for normal work or follow the repository for the next development generation.