# clawbot documents Clawdbot setup for readers who want a relay endpoint

> The clawbot repository is a single Chinese-language README that walks through installing the Clawdbot local AI assistant, running its onboard wizard, pairing a Telegram bot and pointing it at a third-party model provider. It is a guide, not the software, and it steers readers toward one particular relay service.

**xianyu110/clawbot** — Clawdbot完整配置指南：从安装到Claude Code中转

- Repository: https://github.com/xianyu110/clawbot
- Website: https://awesome.tryopenclaw.asia
- Stars: 875 · Forks: 120
- Language: Unknown
- License: not declared
- Published: 2026-10-20 · Updated: 2026-10-20 · Language: en
- Canonical page: https://hysenlabs.com/projects/xianyu110-clawbot

## A repository that contains one README and no code

The first thing to know about xianyu110/clawbot is what it is not. The repository has a single top-level file, README.md. There is no source tree, no test suite and no release history, and the licence field is unknown. The last push was on 2026-09-10, and the described purpose is a configuration guide rather than a piece of software.

What that README is, in practice, is a walkthrough. It opens with a note about a printed book that rewrites the tutorial, and it carries a navigation list of sibling repositories for Claude Code, OpenAI Codex, Gemini image tools and similar subjects. Most of the page is the guide itself, written in Chinese, with sections for system requirements, installation, configuring a Claude Code relay endpoint, verification, common pitfalls, a FAQ and a list of common commands.

That shape matters when you judge it. You will not get a script, a config schema or a supported API from here. You will get one author's path through the setup, including the exact prompts the wizard prints, which is the part that is genuinely hard to reconstruct from memory.

## The onboard wizard is the real installation path

The guide's installation section covers three routes. The recommended one is a global npm install.

```bash
npm install -g clawdbot
```

The second is a shell script fetched over the network and piped straight to bash.

```bash
curl -fsSL https://clawd.bot/install.sh | bash
```

The third builds from a source checkout, which the guide shows as a clone, an install, a build and a link step.

```bash
git clone https://github.com/clawdbot/clawdbot.git
cd clawdbot
pnpm install
pnpm build
npm link
```

Whichever route you take, the guide then runs the same command.

```bash
clawdbot onboard
```

The documented prerequisites are macOS, Linux or Windows, Node.js 22.0.0 or newer, and pnpm or npm as the package manager, with Xcode listed as optional on macOS if you want to build native applications. The first panel of the wizard is a security confirmation that states agents can run commands, read and write files and act through enabled tools, and it asks you to read the security documentation before agreeing.

## ANTHROPIC_BASE_URL is ignored, so providers go in the config file

This is the most useful thing in the guide, because it is the kind of negative knowledge that normally takes an afternoon to discover. The guide states plainly that Clawdbot does not support setting a custom API endpoint through the ANTHROPIC_BASE_URL environment variable, and that the endpoint has to be configured through the providers block of the configuration file instead.

The file lives at ~/.clawdbot/clawdbot.json, and the guide advises backing it up before editing it.

```bash
cp ~/.clawdbot/clawdbot.json ~/.clawdbot/clawdbot.json.bak
nano ~/.clawdbot/clawdbot.json
```

Inside, a models object holds a mode of merge and a providers map. Each provider entry carries a baseUrl, an api field, an apiKey and a list of models with an id, a name, a reasoning flag, an input list, a cost block, a contextWindow and a maxTokens value. The api field is what distinguishes one provider from another: the example uses openai-completions, anthropic-messages and google-generative-ai for the three entries, and the baseUrl paths differ per provider, with a /v1 suffix for one, none for another and /v1beta for the Google-style endpoint.

The guide also lists three authentication choices in the wizard: an Anthropic setup-token, an Anthropic token read from the Claude Code CLI, and an Anthropic API key, with a table comparing who each suits. The setup-token route is described as letting a subscription user reuse subscription quota without buying a separate API key.

## Pairing is a second, separate approval step

Even after the wizard reports success, a Telegram channel is not usable until the owner approves it. The guide describes the sequence: you send a message to the bot, it replies that access is not configured, and it hands you a pairing code.

```
Clawdbot: access not configured.

Your Telegram user id: 1234567890
Pairing code: ABC12345

Ask the bot owner to approve with:
clawdbot pairing approve telegram <code>
```

The approval itself is a command run in the terminal, with the code pasted in place of the placeholder.

```bash
clawdbot pairing approve telegram ABC12345
```

The final panel of onboarding shows the local endpoints, which is worth knowing because the guide also lists a web control panel and a gateway.

```
◇  Control UI ─────────────────────────────────────────────────────╮
│  Web UI: http://127.0.0.1:18789/                                 │
│  Web UI (with token): http://127.0.0.1:18789/?token=your-token   │
│  Gateway WS: ws://127.0.0.1:18789                                │
├──────────────────────────────────────────────────────────────────╯

└  Onboarding complete.
```

The panel also shows the channel status for Telegram, WhatsApp and Discord, and the guide says a dozen-plus platforms are supported. The wizard reports a heartbeat interval of one hour for the main agent, and the channel list in the guide says Telegram, WhatsApp, Discord and others are all optional.

## The relay endpoint is the guide's own commercial choice

A reader should notice what the guide is pointing at. The provider examples all use the same host, tryallapi.com, with a base URL of https://tryallapi.com and a key shaped like sk-xxxxxxxxxxxxx. The guide then lists that service as a recommended relay and adds a purchase link on another domain. The book notice at the top and the list of sibling tutorial sites reinforce the pattern: this is a guide written to funnel readers toward a set of paid services.

That does not make the technical content wrong. The mechanics of the providers block, the fact that ANTHROPIC_BASE_URL is not honoured, and the pairing handshake are all facts about how the tool behaves, and they are the parts a reader cannot easily guess.

It does mean two things are missing. The guide does not explain how a relay operator handles the credentials it is given, and it does not compare the cost of a relay against the direct provider pricing that the wizard itself offers as the third authentication method. If you follow the guide with your own Anthropic API key instead of a relay, the provider block is where you point it at your own endpoint.

## The guide is thin exactly where you would want documentation

There are no GitHub releases, so there is no version of the guide to pin, and the repository carries no licence field, so the reuse terms of the text are not stated. The top-level file list is README.md alone, which means there is no changelog, no issue history and no automation to speak of.

The guide also points to a separate, newer deployment tutorial hosted on another domain, and describes the current page as the earlier one. For a reader, that is a fork in the road: two documents, both Chinese, covering overlapping ground, with no stated rule about which one to follow when they disagree.

The verification section the table of contents promises is largely a list of things to check in the wizard output, such as whether the channel reports ok and whether the control panel address responds, and the pitfalls section is a set of short notes rather than a catalog of failure modes. For a tool whose agents can run commands and write files, the guide's own security warning is the most important line it contains, and it is the one line that points elsewhere, to the security documentation at docs.clawd.bot.

## Conclusion

clawbot is worth reading if you are setting up Clawdbot for the first time in Chinese and want a worked path through the onboard wizard, the pairing handshake and the providers block in clawdbot.json, since the sequence it records is specific and hard to guess. It is not a place to look for a maintained configuration tool, because the repository holds a README and nothing else, and it says nothing about the security of the third-party relay it points at. Verify first which endpoint and credential you are about to paste into ~/.clawdbot/clawdbot.json, and confirm that the account behind that key is yours and not a shared one.

## FAQ

### What does the clawbot repository contain?

It contains a single README.md, a Chinese-language configuration guide for the Clawdbot local AI assistant. There is no source code, no test suite and no GitHub release, and the licence field is unknown.

### How do I install Clawdbot following the clawbot guide?

The guide gives three routes: npm install -g clawdbot as the recommended one, a curl one-liner that pipes the install script to bash, and a source build from a git clone with pnpm install, pnpm build and npm link. Node.js 22.0.0 or newer is required.

### Can I set a custom API endpoint with ANTHROPIC_BASE_URL in Clawdbot?

No. The guide states that Clawdbot does not support the ANTHROPIC_BASE_URL environment variable for that purpose, and that the endpoint must be set in the providers block of ~/.clawdbot/clawdbot.json.

### How do I approve a Telegram bot in Clawdbot?

Send the bot a message, it replies with a pairing code, and the owner approves it from the terminal with clawdbot pairing approve telegram followed by that code.

### Where does the clawbot guide send its readers for API access?

Its provider examples all use the host tryallapi.com, and it lists that service as a recommended relay with a purchase link. The guide itself also offers an Anthropic API key as one of three authentication methods in the onboarding wizard.

## Sources

- [Issues](https://github.com/xianyu110/clawbot/issues)
- [Project website](https://awesome.tryopenclaw.asia)
- [README](https://github.com/xianyu110/clawbot/blob/main/README.md)
- [xianyu110/clawbot on GitHub](https://github.com/xianyu110/clawbot)

---

Hysen Labs editorial analysis, written from the project's own repository and release notes. Cite the canonical page: https://hysenlabs.com/projects/xianyu110-clawbot
