Sharpwire is a desktop playground for building and running multi-agent workflows visually.
It is intentionally not a production platform. The goal is to make it easy (and fun) to experiment with agent behavior, tool wiring, orchestration patterns, and plugin-driven extension without heavy setup.
- Build agent systems quickly with a visual graph instead of stitching everything together by hand.
- Try ideas fast: tweak prompts, tools, and handoff paths, then run again.
- Watch agent behavior in one place (chat, graph state, and logs).
- Explore orchestration patterns (explicit wired flows + orchestrator-driven delegation).
- Use it as a safe sandbox for learning, prototyping, and debugging agent concepts.
- Visual agent graph: connect agents with default and return/feedback handoffs.
- Chat-driven execution: send a task and let the orchestrator route work.
- Tool-enabled agents: agents can call tools and coordinate through handoffs.
- Plugin system: load workspace plugins and expose plugin settings in-app.
- Session/workspace state: persist scenes, agent layouts, and local settings.
- Windows installer + auto-updates: builds are published on GitHub Releases; the installed app can check for newer versions and apply updates (see Settings → Updates).
One of the most interesting parts of Sharpwire is that it can self-extend:
- You can add new tools and behaviors via plugins.
- Agents can use existing tools to create or update assets in the workspace.
- The orchestrator can incorporate new capabilities as they become available.
In short: Sharpwire can grow its own toolbox while you experiment, which makes it great for agentic prototyping and "what if?" workflows.
- Developers curious about multi-agent systems.
- People experimenting with orchestration, tools, and prompt design.
- Anyone who wants a local, visual way to play with agents for fun.
If you need hardened reliability, strict security boundaries, or enterprise guarantees, this project is not aiming for that today.
Sharpwire is a Windows desktop app. The fastest way to try it is the latest installer from GitHub Releases.
- From that page, download the latest Windows installer (the setup
.exefrom the newest release). - Run the installer and launch Sharpwire.
- Open Settings and add the LLM provider API keys you plan to use.
- Use the default workspace folder (or pick another) and open the Chat tab.
- Send a prompt (see Example prompts below). Watch the graph, chat, and agent logs as the orchestrator delegates work.
Use these as copy-paste starters once keys are configured. They are meant to show delegation, tools, and multi-step behavior—not to be “perfect” tasks.
- Simple handoff — “Ask the Coder to add a small
hello.pyin the workspace that prints a classic Python hello-world line, then ask the Reviewer to skim the script and confirm it is reasonable.” - Team of agents — “Create a team of agents that can brainstorm, implement, and market a text-based Python game.”
- Iterate on a file — “Create
notes.mdwith three bullet ideas for a weekend project, then improve the wording and save the result.” - Explore the graph — “Explain what agents you see on the graph and what you would delegate to each one for a ‘parse a CSV and suggest column types’ task—don’t run tools yet, just the plan.”
After each run, tweak the graph (agents, wiring, tools) and send the same prompt again to compare behavior.
Use this path when you want to change Sharpwire itself or run from source.
Prerequisites
- Windows with .NET 10 SDK installed.
Run from the repository
git clone https://github.com/distantdev/sharpwire.git
cd sharpwire
dotnet build
dotnet run --project Sharpwire.csprojThen configure Settings (API keys) the same way as the installed build. For how releases are built and published, see RELEASES.md.
Sharpwire is about rapid iteration, playful exploration, and learning by building agent systems hands-on.