Mix local Ollama, OpenAI-compatible endpoints, and OpenAI in one graph.
Visual. Local-first. Open source.
Design, run, and inspect multi-agent workflows on your own machine.
Build orchestration graphs with a React Flow canvas, mix providers per agent, stream traces live, and keep control over your models, keys, and runtime.
Local-first
Ollama + OpenAI-Compatible + OpenAI
Visual orchestration
Bring your own providers
CEO
coordinator · local ollama
Plans the workflow, delegates to specialists, and merges outputs.
Researcher
openai-compatible
Finds context and produces concise upstream briefs.
Developer
openai
Transforms the brief into implementation-ready output.
Use the one-line installer instead of cloning and wiring the project manually.
Track node-by-node events, outputs, and execution state from a single trace view.
Install
Get running from a single command.
The installer downloads a versioned release artifact from GitHub and sets up a local launcher. After install, start the studio and configure your providers from the UI.
curl -fsSL https://raw.githubusercontent.com/harishkotra/local-agent-studio/main/install.sh | bashInstall
Run the curl command to fetch and install the latest packaged release.
Launch
Start the studio locally and open the app in your browser.
Configure
Add your provider keys or local Ollama instance, then build your orchestration.
Why This Exists
An orchestration tool that stays out of your way.
Bring your own stack
Users decide which provider and model each agent should run on. There is no forced platform lock-in.
Local by default
Keys stay on the user’s machine, workflows are portable, and the default operating mode does not require a hosted service.
Built for visual reasoning
Agents, routers, inputs, HTTP tools, and outputs are all composed on a graph users can actually inspect.
Open to extension
The long-term direction is provider and skill compatibility that fits how the ecosystem already works.
How It Works
Three steps from blank canvas to traceable workflow.
1
Create providers and profiles
Add Ollama, OpenAI-compatible, or OpenAI providers, then define reusable agent profiles with their own prompts, roles, notes, and models.
2
Connect your orchestration graph
Use the React Flow canvas to place agents, add inputs and outputs, and wire the handoff path visually.
3
Run and inspect
Execute the workflow locally, stream node events in real time, and inspect the full trace after completion.
Compatibility
Built for the way people already run models.
Supported today
- Ollama for local model execution
- OpenAI-compatible APIs such as Featherless.ai
- OpenAI models through user-provided keys
Planned next
- AgentSkills compatibility for ecosystem-aligned capabilities
- Richer tool builders and workflow inputs
- Release and installer polish for broader distribution
Open Source
Use it locally. Fork it. Improve it.
The product is open source and built in public. The repo has issues, roadmap work, and installation docs so contributors can get involved quickly.
FAQ
Common questions before you install.
Is this local-first?
Yes. The MVP is designed for single-user local execution and local credential storage.
Do I need OpenAI?
No. You can run with Ollama or any OpenAI-compatible provider instead.
Can different agents use different providers?
Yes. Each agent profile can point to its own provider and model configuration.
Where do I contribute?
Contributions happen in the main project repository on GitHub through issues and pull requests.