MCP Servers · thClaws

thClaws

thClaws is an open-source AI agent harness written in Rust that runs on desktop, CLI, and web from a single binary. It supports multiple LLM providers, integrates with the Model Context Protocol for extensibility, and includes skills, plugins, and agent orchestration features.

Talk to DEV.co Discuss Implementation

Source: GitHub — github.com/thClaws/thClaws

1.1k

GitHub stars

158

Forks

Rust

Primary language

Apache-2.0

License (OSI-approved)

Key facts

Objective fields from the source. Values we can't verify are shown as “Unknown” rather than guessed.

Field	Value
Repository	thClaws/thClaws
Owner	thClaws
Primary language	Rust
License	Apache-2.0 — OSI-approved
Stars	1.1k
Forks	158
Open issues	1
Latest release	v0.87.0 (2026-07-05)
Last updated	2026-07-05
Source	https://github.com/thClaws/thClaws

What thClaws is

A Rust-native agent framework with multi-provider LLM support (Anthropic, OpenAI, Gemini, Ollama, etc.), MCP-based tool integration, JSONL-based session persistence, filesystem sandbox, and a shared agent loop across desktop (Tauri), CLI REPL, non-interactive, and WebSocket/HTTP webapp surfaces.

Quickstart

Get the thClaws source

Clone the repository and explore it locally.

terminalbash

1git clone https://github.com/thClaws/thClaws.git2cd thClaws3# follow the project's README for install & configuration

Need it deployed, integrated, or customized instead? DEV.co ships production installs.

Best use cases

Multi-model AI development workflows

Teams building agent-driven applications can leverage one binary across desktop, CI/headless, and web—switching LLM providers mid-session and composing tools via MCP without language lock-in.

Local-first automation and coding assistants

Organizations requiring sovereign AI (on-machine, no external APIs mandatory) can use Ollama/LMStudio backends, filesystem sandboxing, and scheduled agent loops for overnight builds or recurring tasks.

Extensible skill and plugin ecosystems

Projects needing custom AI workflows can ship domain logic as YAML-frontmatter skills, MCP servers, or shell hooks without recompiling, enabling rapid iteration and community contributions.

Implementation considerations

Verify Rust/Tauri build chain for target platforms (macOS, Windows, Linux) and plan CI/CD for binary distribution.
API key management defaults to OS keychains; audit Secret Service/Credential Manager integrations and `.env` fallback for your security model.
Session JSONL storage and `.thclaws/` project directory structure require git-ignore rules to avoid committing API calls or large chat logs.
MCP server stdio/HTTP integration adds dependency management complexity; plan vetting and lifecycle (startup/shutdown) for third-party servers.
Agent orchestration depth (3 levels of subagents, concurrent side-channels) scales token usage and latency quickly; monitor token budgets and tool-call limits per model.

When to avoid it — and what to weigh

Mature production SLA required — Project is ~3 months old (April 2026 start date). Single-binary architecture and rapid release cadence signal ongoing breaking changes; production use requires tolerance for instability.
Enterprise Java/C# ecosystem — thClaws is Rust-only. Integration with legacy Java EE, C#/.NET, or COBOL systems requires custom adapters or API wrappers—not a primary use case.
Fully air-gapped or regulated compliance mandates — Default Keychain/Secret Service integration, cloud provider support (Azure, DashScope, OpenRouter), and webhook/Telegram integrations complicate purely offline deployments; requires custom configuration.
Low-code/no-code non-technical teams — Configuration (`.thclaws/settings.json`, YAML skills, MCP setup) assumes CLI/JSON fluency. UI is functional but minimal; not a point-and-click SaaS replacement.

License & commercial use

Licensed under Apache License 2.0 (permissive OSI license). README badge mentions "MIT OR Apache-2.0"; verify actual LICENSE file to confirm dual-license or Apache-2.0 only. Allows commercial use, modification, and distribution with liability/trademark caveats.

Apache-2.0 is permissive and allows commercial use, including proprietary modifications and distribution. No explicit restrictions on enterprise or SaaS use. However, verify the actual LICENSE file against the README badge claim of dual MIT/Apache; if Apache-2.0 only, no additional restriction beyond standard warranty/liability disclaimers.

DEV.co evaluation signals

Editorial assessment — not user reviews. Directional, with an explicit confidence level.

Signal	Assessment
Maintenance	Active
Documentation	Adequate
License clarity	Clear
Deployment complexity	Moderate
DEV.co fit	Good
Assessment confidence	High

Security considerations

Filesystem sandbox scoped to working directory and destructive command flagging provide baseline containment. OS keychain integration (macOS, Windows, Linux Secret Service) is standard. No explicit mention of input validation, prompt injection hardening, or tool-call signature verification. Audit required for production deployments involving untrusted user input or multi-tenant scenarios. Telegram/web integrations expose agent to internet; network policies and rate-limiting not documented.

Alternatives to consider

Anthropic Claude Code (web + SDK)

Official, mature, multi-model reasoning; closed-source and cloud-only. Better for teams already in Anthropic ecosystem and accepting vendor lock-in.

OpenAI Assistants API + custom UI

Simpler model access, structured tool calling, hosted. Lacks the local-first, multi-provider, and extensibility (MCP, skills) of thClaws.

LangChain / LlamaIndex frameworks

Python-based, mature ecosystem, broad LLM/data integration. Requires more orchestration code; thClaws bundles batteries-included agent loop and desktop UI.

Software development agency

Build on thClaws with DEV.co software developers

Download the latest release, review the manual, and join GitHub discussions to test multi-provider switching, skill development, and MCP integrations. Contribute agent backends or connectors to shape v1.0.

Talk to DEV.co

Related open-source tools

Surfaced by semantic similarity across the DEV.co open-source index.

webclaw onecli goose

Related on DEV.co

Explore the category and the services that help you build with it.

MCP Servers AI Development Custom Software & Automation API Development Talk to DEV.co

thClaws FAQ

Can I run thClaws entirely offline?

Yes, using Ollama or LMStudio as the LLM provider and disabling cloud integrations (DashScope, OpenRouter, Telegram, etc.). Keychain fallback to `.env` for local API keys. Some skills or MCP servers may require internet; verify per feature.

Is thClaws production-ready?

Not yet. v0.87.0 is pre-v1.0; rapid release cadence and small team suggest active development and potential breaking changes. Suitable for pilots, internal tools, and community feedback; commercial SLA deployments require risk acceptance.

How do I extend thClaws with custom tools?

Three routes: (1) MCP servers (stdio/HTTP), (2) skills (YAML + shell scripts), (3) plugins (skills + commands + agent defs + MCP bundled). No Rust recompile needed for most extensions. Lifecycle hooks (`pre_tool_use`, etc.) enable shell-script integrations.

What LLM providers does thClaws support?

Anthropic (native + Claude Code SDK), OpenAI, Google (Gemini, Gemma), Alibaba (DashScope, Qwen), DeepSeek, Z.ai, NVIDIA NIM, NSTDA Thai models, OpenRouter, Ollama, LMStudio, plus generic OpenAI-compatible (LiteLLM, Portkey, etc.). Switch mid-session via `/model` or `/provider`.

Software development & web development with DEV.co

From first prototype to production, DEV.co delivers software development services around tools like thClaws. Our software development agency staffs experienced software developers and web developers for custom software development, web development, integrations, and ongoing support across mcp servers and beyond.

Evaluate thClaws for Your AI Agent Workflow

Talk to DEV.co Schedule a Consultation