Thanks to visit codestin.com
Credit goes to thclaws.ai

new · v0.126.0 many agents, one workspace
[ open-source agent harness ]

your machine. your agent.

thClaws is a native-Rust AI agent workspace that codes, automates, remembers, and coordinates — running locally on your machine. multi-provider. open standards. no cloud lock-in.

07 platforms
07+ providers
01 single binary
your data, your machine
~/projects/landing   —   thclaws
$ thclaws --cli
· loaded settings from .thclaws/settings.json
· provider: anthropic · model: claude-sonnet-4-6
read README.md and tell me what's missing
◌ Read README.md (3,287 tokens)
◌ thinking…
▍ a Roadmap section, a Threat Model note, and the keybinding
cheat-sheet you mentioned in #14. want me to draft them?
§ 01

see it work

Three tabs. One binary. Captured live from a session of thClaws looking at this very site.

  1. thClaws Files tab — landing-page source open in the editor with file tree
  2. thClaws Terminal tab — REPL session with ASCII banner
  3. thClaws Chat tab — conversational interface answering a product question
§ new

what's new

A workspace is no longer one agent — it is a shelf of them. Plus media generation and multi-model deliberation, built in.

v0.126.0 · multiple agents

many agents, one workspace

Keep a coding agent, a research agent and a writing agent side by side in the same project, and switch between them from a rail on the left — each keeps its own conversation, scroll position and half-typed message while you are away. Every agent lives in its own folder, with its own sessions, memory, settings and browser logins, running as its own process. Get one from Agent Templates or start a blank agent; up to 8 run at once, in the desktop app and in --serve.

Existing workspaces upgrade themselves the first time v0.126.0 opens them: your project moves into .thclaws/bots/main/, git history included, and thclaws -p at the root still reaches it. Anything holding the old path — an editor, a script — needs the new one.

thClaws Media Studio — text-to-image with a live gallery of generated media
media studio

generate images & video, built in

Text → Image, Image Edit, Text → Video, Image → Video — across Gemini, OpenAI gpt-image-2, Qwen, Veo, and DashScope HappyHorse. Pick a model, set the resolution, generate; results drop into a live gallery you can click to reuse as the next source. The same tools are callable by the agent from chat.

thClaws — configuring the OpenRouter Fusion panel, judge, and limits
openrouter fusion

many models, one answer

openrouter/fusion fans a question across a panel of up to 8 models, then a judge synthesizes the consensus. New openrouter/fusion+ makes the panel, judge, tool-call budget, temperature, and reasoning fully configurable straight from the model picker.

§ 02

three interfaces
one binary

The same engine, the same sessions, the same config — dressed for whichever surface you prefer. Switch freely; your history follows you. Add a fourth via thclaws --serve when you need WebSocket access from a tunneled remote machine.

A.

Desktop GUI

$ thclaws
terminal chat files team
refactor src/parser.rs
◌ Edit · src/parser.rs · +42 / -18
◌ Bash · cargo test
▍ all 47 tests pass. parser is 22% smaller.

Native window. Terminal, Chat, Files, and optional Team tabs. Theme-aware (light / dark / system).

B.

CLI REPL

$ thclaws --cli
$ thclaws --cli
thclaws v0.8.4 · type /help
/provider openai
✓ provider → openai (gpt-4.1)
! ls -1 | head

Interactive prompt for SSH, headless servers, or zero-GUI sessions. Every slash command works identically.

C.

One-shot

$ thclaws -p «…»
$ thclaws -p "summarise CHANGELOG" -v
◌ Read CHANGELOG.md
▍ v0.8.x ships /agent side-channels, /dream
KMS consolidator, scheduling daemon, hooks.
[tokens: 1,247in/342out · 4.1s]
$

Single turn, then exits. Compose with other tools, drop into CI pipelines, or fire from a shell one-liner.

§ 03

capabilities

Not a chat wrapper. A full agent harness with the surface area to actually get work done — and the boundaries to do it safely.

§ 04

install

Pre-built binaries for six targets plus an official Docker image. Or build from source if that's more your speed.

# Recommended — universal .dmg (Apple Silicon + Intel in one file)
curl -L -o thclaws.dmg https://github.com/thClaws/thClaws/releases/download/v0.130.0/thclaws-v0.130.0-universal-apple-darwin.dmg
open thclaws.dmg            # drag thClaws into Applications

# Or tarball (headless / SSH boxes)
# Apple Silicon:
curl -L https://github.com/thClaws/thClaws/releases/download/v0.130.0/thclaws-v0.130.0-aarch64-apple-darwin.tar.gz \
  | tar -xz && sudo mv thclaws /usr/local/bin/

# Intel:
curl -L https://github.com/thClaws/thClaws/releases/download/v0.130.0/thclaws-v0.130.0-x86_64-apple-darwin.tar.gz \
  | tar -xz && sudo mv thclaws /usr/local/bin/

Browse all assets & SHA-256 sidecars on the downloads page or the github releases tab.

§ 05

three commitments

  1. 01.

    local first, sovereign by design

    Your code, your prompts, your conversations stay on your machine. Files in your filesystem; sessions on your disk; keys in your OS keychain. Cloud is opt-in per request, never the default.

  2. 02.

    no provider lock-in

    Switch providers mid-session. Run an Anthropic model in the morning, an Ollama model on a flight, an OpenAI model when you need a one-shot. The harness doesn't care; your prompts and skills come with you.

  3. 03.

    open standards, not a walled garden

    Built on the conventions the agent-tooling industry is converging on. MCP. AGENTS.md. SKILL.md. Your config is portable to every harness that speaks the same standards — and to whatever comes next.

get started in < 60 seconds

One command. No account. No telemetry phone-home. The binary runs, you tell it what to do.