Practical overview
When it fits
Use Leagent when you need a hackable desktop agent stack that combines local tools, visual workflow construction and model-provider flexibility in one deployment.
Who it is for
- Developers building custom AI agents and tool-using workflows.
- Teams seeking a self-hosted desktop environment for document, research, data or coding tasks.
- Users working with local models through Ollama or vLLM.
Documented uses
- Plan and execute multi-step tasks through tool calls.
- Build and run visual workflows whose tools appear as typed nodes.
- Extract, transform and summarise documents and research papers.
- Analyse CSV, Excel and database data and produce charts or generated interfaces.
- Create coding-project workflows and offline game-art asset pipelines.
Requirements & limits
The repository identifies Leagent as open source under the Apache-2.0 licence. It documents Python 3.11+, React 19, SQLite or PostgreSQL, with SQLite and a single process as the default deployment path.
- Local execution is supported through Ollama or vLLM, while remote providers including OpenAI, Anthropic, Azure OpenAI, DeepSeek and DashScope are also documented.
- The built-in tool catalogue spans documents, web research, data, code, databases, media and generative UI.
- ReactFlow workflows provide typed nodes, YAML export, reusable templates, concurrent independent branches and central retry, timeout and backoff handling.
- PostgreSQL and Milvus are documented as scale-out options beyond the SQLite default.
- Research Paper Mode and the game-art pipeline are specialised capabilities in addition to general agent workflows.
Check current requirements in the official documentation. Source-based guidance does not mean OpenViST has run or independently tested this project.
Set up LeAgent for local development
Prepare and start LeAgent’s local backend and frontend development environment.
Before you start
- Git
- uv
- Node.js 20+ or 22+
Steps
Clone the LeAgent repository and change into its directory.
Run
./start.shto start the backend on port 7860 and frontend on port 5173.Set at least one provider key, either as an environment variable or through Settings → Environment secrets in the web UI.
Expected outcome
The local development services start, with a provider key configured for model use.
Guide limitations
- The start script synchronizes the Python environment, installs frontend dependencies, and normally installs Playwright Chromium for web tools.
- At least one provider key is required; provider setup is separate from starting the development services.
Set up LeAgent with Docker Compose
Build and start LeAgent’s Docker deployment with its default SQLite-backed container.
Before you start
- Docker Compose
- A provider key
- A value for
LEAGENT_SECRET_KEY
Steps
Change into the repository’s
deploydirectory.Copy
.env.exampleto.envand setLEAGENT_SECRET_KEYand at least one provider key.Run the command below.
docker compose up -d --buildOpen
http://localhost:8000/docsfor the API and interactive documentation.
Expected outcome
The Docker deployment starts, and API documentation is available at the documented local URL.
Guide limitations
- The default image uses SQLite in a single container.
- A local GPU vLLM service is described as an optional compose overlay; its setup steps are not included here.
Sources & review
Source record details
- Source availability
- Active
- Sources checked
- 3 October 2026
- Original website title
- vixues/LeAgent · Official homepage title · retained in its source language