A .NET multi-agent orchestration framework built on Microsoft Agent Framework. Define a team of AI agents in a YAML config and coordinate them via keyword routing, state machine routing, declarative directed-graph routing, LLM-based selection, or fully autonomous Magentic orchestration. Works with Anthropic, xAI, OpenAI, Azure OpenAI, Ollama, and any OpenAI-compatible provider. Agents can be local or remote — the A2A protocol lets you federate agent slots to independently hosted services.
Early stage. Functional but not battle-tested — best suited for experimentation and automation of well-defined tasks.
Pipelines range from a single task-routed assistant:
...to multi-agent workflows with conditional keyword routing and anti-hallucination validators enforced at every handoff:
...to declarative directed-graph pipelines where back-edges express review cycles without duplicating states:
...to parallel fan-out/fan-in where a coordinator spawns concurrent workers that merge into a single downstream node:
...to fully autonomous Magentic orchestration where a Manager dynamically selects agents and collects their reports:
- .NET 10 SDK
- An API key for your chosen provider (e.g.
XAI_API_KEYfor xAI)
# Build
./build.sh # Linux / macOS
.\build.ps1 # Windows
# Generate a config from a template (interactive wizard)
./bin/fuseraft init
# Or pick a template directly
./bin/fuseraft init --template dev-team --model claude-sonnet-4-6
./bin/fuseraft init --template graph # directed-graph pipeline with parallel fan-out
./bin/fuseraft init --template designer # AI-assisted config designer
# Run against a YAML config
./bin/fuseraft run -c config/orchestration.yaml "Build a REST API in Go with JWT authentication"
# Resume the most recent incomplete session
./bin/fuseraft run --resume
# Validate a config before running; --diagram prints a Mermaid flowchart
./bin/fuseraft validate config/orchestration.yaml
./bin/fuseraft validate config/orchestration.yaml --diagram- Coordinates any number of agents via keyword routing, state machine routing, declarative directed-graph routing (with parallel fan-out/fan-in), LLM-based selection, or fully autonomous Magentic orchestration
- Agents can be declared inline or as standalone
AgentFileYAML files — reuse and version agent definitions independently across configs - Gives each agent access to tools: filesystem, shell, git, HTTP, JSON, search, Docker sandboxes, MCP servers, persistent scratchpad, and a shared chatroom
- Federates agent slots to remote services via the A2A protocol — remote agents participate identically to local ones
- Checkpoints after every turn so sessions can always be resumed exactly where they left off
- Tracks token usage per turn; enforces per-model context caps via
MaxContextTokensand a session-wide hard stop viaMaxTotalTokens - Blocks handoffs with routing validators unless evidence of real progress is present on disk
- Sandboxes agent file and shell access to a configured directory tree
- Applies per-agent execution rings, prompt injection detection, circuit breaker, and a hash-chain audit log via the Agent Governance Toolkit
- Supports mixing any combination of LLM providers per agent
- Streams agent turns to a browser-based DevUI (
--devui) for real-time session visualization - Includes an interactive Orchestration Designer (
fuseraft init --template designer) — describe your use case and get a validated config back
| Doc | What it covers |
|---|---|
| Getting Started | Prerequisites, build, first run |
| CLI Reference | All commands and flags |
| Configuration | Full config schema (YAML and JSON) |
| Models & Providers | Model configuration and provider auto-detection |
| Plugins | All built-in tools agents can call |
| Strategies | Selection and termination strategies |
| Routing Validators | Anti-hallucination handoff guards |
| Harness Engineering | Designing orchestration configs that enforce real progress mechanically |
| MCP Integration | Connecting external MCP servers |
| Security & Sandbox | File and network containment |
| Governance | Execution rings, audit log, circuit breaker, SLO tracking |
| Context Store | Importing files and directories into the session context |
| Sessions | Resumption, HITL, cost tracking, compaction |
| Examples | Ready-to-use config examples |
| Design | Architecture, layer map, MAF usage, and decision log |
The Fuseraft VS Code extension brings the CLI into your editor:
- Activity bar panel with Run Task, Sessions, Configs, and Context views
- CodeLens on config files — run, validate, or diagram without leaving the editor
- Command palette integration for all common CLI commands
- YAML/JSON IntelliSense with schema validation for config files
- Session transcript viewer and real-time DevUI launcher
MIT