SharpClaw Code

OpenCode meets Claude Code in C#.

SharpClaw Code is a .NET-native coding agent runtime for teams who want modern agent ergonomics, explicit orchestration, and production-grade architecture in the C# ecosystem.

It is designed for cross-platform use, strong Windows compatibility, permission-aware tooling infrastructure, durable sessions, MCP and plugin extensibility, and machine-readable CLI workflows that scale beyond a toy prototype.

Why SharpClaw Code

• C# and .NET first: built around .NET 10, C# 13, idiomatic dependency injection, and System.Text.Json.
• Agent-runtime ready: durable sessions, append-only history, runtime lifecycle control, and structured telemetry.
• Tooling with guardrails: explicit permission modes, mediated tool execution paths, and clear operational boundaries.
• Extensible by design: model providers, MCP servers, plugins, and skills can plug into the same runtime surface.
• Built for real workflows: REPL, one-shot prompts, diagnostics, status inspection, and automation-friendly CLI behavior.

Requirements

• .NET SDK 10
• C# 13
• Optional: GitHub CLI (gh) for releases and repository automation

Quick Start

git clone https://github.com/clawdotnet/SharpClawCode.git
cd SharpClawCode
dotnet build SharpClawCode.sln
dotnet test SharpClawCode.sln

Run the CLI from the repository root:

dotnet run --project src/SharpClaw.Code.Cli -- [arguments]

CLI Examples

# Interactive REPL (default when no subcommand)
dotnet run --project src/SharpClaw.Code.Cli
dotnet run --project src/SharpClaw.Code.Cli -- repl

# One-shot prompt
dotnet run --project src/SharpClaw.Code.Cli -- prompt "Summarize this workspace"

# Operational diagnostics
dotnet run --project src/SharpClaw.Code.Cli -- doctor
dotnet run --project src/SharpClaw.Code.Cli -- status

# Machine-readable output
dotnet run --project src/SharpClaw.Code.Cli -- --output-format json doctor

Built-in slash commands in the REPL include /help, /status, /doctor, /session, /commands, /mode, /editor, /export, /undo, /redo, and /version. Use /help to see the current built-in list plus any workspace-defined custom commands, and /exit to leave interactive mode.

What You Get

Capability	What it means
Durable sessions	Persist conversation state, turn history, and recovery data for longer-running agent work
Permission-aware tools	Built-in and plugin-backed tools run through explicit permission policies instead of hidden side effects
Provider abstraction	Work with Anthropic and OpenAI-compatible backends through a typed runtime surface
MCP and plugins	Extend the runtime with MCP servers, plugin discovery, and lifecycle-aware integrations
Structured output	Support human-readable CLI flows and JSON-friendly automation from the same commands

Solution Layout

Area	Project(s)
CLI	SharpClaw.Code.Cli, SharpClaw.Code.Commands
Core contracts	SharpClaw.Code.Protocol
Runtime and orchestration	SharpClaw.Code.Runtime
Agents	SharpClaw.Code.Agents
Tools and permissions	SharpClaw.Code.Tools, SharpClaw.Code.Permissions
ACP host	SharpClaw.Code.Acp
Providers, MCP, and plugins	SharpClaw.Code.Providers, SharpClaw.Code.Mcp, SharpClaw.Code.Plugins
Memory, git, web, and skills	SharpClaw.Code.Memory, SharpClaw.Code.Git, SharpClaw.Code.Web, SharpClaw.Code.Skills
Sessions, telemetry, and infrastructure	SharpClaw.Code.Sessions, SharpClaw.Code.Telemetry, SharpClaw.Code.Infrastructure

For dependency boundaries and project responsibilities, see docs/architecture.md and AGENTS.md.

Global CLI Options

Option	Description
--cwd <path>	Working directory; defaults to the current directory
--model <id>	Model id or alias; provider/model forms are supported where configured
--permission-mode <mode>	readOnly, workspaceWrite, or dangerFullAccess; see docs/permissions.md
--output-format text|json	Human-readable or structured output
--primary-mode <mode>	Workflow bias for prompts: build or plan
--session <id>	Reuse a specific SharpClaw session id for prompt execution

Subcommands include prompt, repl, doctor, status, session, commands, mcp, plugins, acp, bridge, and version. Workspace custom commands can also appear as top-level commands when discovered.

Supported --output-format values today are text and json. Unknown global option values currently fall back to defaults, so scripts should pass exact values.

Documentation

Doc	Topic
docs/architecture.md	Solution layout, dependencies, and execution flows
docs/runtime.md	Turns, ConversationRuntime, and runtime coordinators
docs/sessions.md	Session snapshots and append-only event logs
docs/providers.md	IModelProvider, configuration, and provider integration
docs/tools.md	Tool registry, executor behavior, and built-ins
docs/permissions.md	Permission modes, policies, and approvals
docs/agents.md	Agent Framework integration
docs/mcp.md	MCP registration, lifecycle, and orchestration
docs/acp.md	ACP stdio host and protocol notes
docs/plugins.md	Plugin discovery, trust, and CLI flows
docs/testing.md	Unit, integration, and parity-harness coverage
docs/migrations-deferred.md	Deferred migrations and follow-up work
ARCHITECTURE-NOTES.md	Architectural follow-ups and future cleanup ideas

Current Scope Notes

• The shared tooling layer is permission-aware and available across the runtime, but the current Microsoft Agent Framework bridge still focuses on provider-backed runs rather than a full tool-calling loop inside the chat framework path.
• Operational commands support stable JSON output envelopes via --output-format json, making them suitable for scripting and automation.

Contributing

Use .NET 10 and C# 13. Follow AGENTS.md for project boundaries, JSON conventions (System.Text.Json only), logging, and cross-platform behavior. Before opening a PR, make sure the solution builds and the relevant tests pass:

dotnet build SharpClawCode.sln
dotnet test SharpClawCode.sln

Repository: github.com/clawdotnet/SharpClawCode