Claw Code Agent
A Python reimplementation of the Claude Code agent architecture โ local models, full control.
---
## ๐ About
This repository reimplements the [Claude Code](https://docs.anthropic.com/en/docs/claude-code) npm agent architecture **entirely in Python**, designed to run with **local open-source models** via an OpenAI-compatible API server.
Built on the public porting workspace from [instructkr/claw-code](https://github.com/instructkr/claw-code), the active development lives at [HarnessLab/claw-code-agent](https://github.com/HarnessLab/claw-code-agent).
> **Goal:** Not to ship the original npm source, but to reimplement the full agent flow in Python โ prompt assembly, context building, slash commands, tool calling, session persistence, and local model execution.
---
## โจ Key Features
| Feature | Description |
|---------|-------------|
| ๐ค **Agent Loop** | Full agentic coding loop with tool calling and iterative reasoning |
| ๐งฐ **Core Tools** | File read / write / edit, glob search, grep search, shell execution |
| ๐ฌ **Slash Commands** | Local commands: `/help`, `/context`, `/tools`, `/memory`, `/status`, `/model`, and more |
| ๐ง **Context Engine** | Automatic context building with CLAUDE.md discovery and usage reporting |
| ๐ **Session Persistence** | Save and resume agent sessions across runs |
| ๐ **Permission System** | Granular control: `--allow-write`, `--allow-shell`, `--unsafe` |
| ๐๏ธ **OpenAI-Compatible Runtime** | Python client targets an OpenAI-compatible API, with `vLLM` as the documented setup |
| ๐ **Qwen3-Coder** | First-class support for `Qwen3-Coder-30B-A3B-Instruct` via vLLM |
---
## ๐ Roadmap
### Testing
- See [TESTING_GUIDE.md](TESTING_GUIDE.md) for concrete commands to verify the current implementation feature by feature.
### Done
- [x] Python CLI agent loop
- [x] OpenAI-compatible local model backend
- [x] Qwen3-Coder support through vLLM with `qwen3_xml` tool parser
- [x] Core tools: `list_dir`, `read_file`, `write_file`, `edit_file`, `glob_search`, `grep_search`, `bash`
- [x] Context building and `/context`-style usage reporting
- [x] Slash commands: `/help`, `/context`, `/context-raw`, `/prompt`, `/permissions`, `/model`, `/tools`, `/memory`, `/status`, `/clear`
- [x] Session persistence and `agent-resume` flow
- [x] Permission system (read-only, write, shell, unsafe tiers)
- [x] Unit tests for the Python runtime
- [x] `pyproject.toml` packaging with `setuptools`
### In Progress
- [ ] Full MCP support
- [ ] Full plugin system
- [ ] Full slash-command parity
- [ ] Full interactive REPL / TUI behavior
- [ ] Exact tokenizer / context accounting
- [ ] Hooks parity
- [ ] Remote modes parity
- [ ] Voice / VIM parity
- [ ] Some deeper runtime details from the npm source
- [ ] Cost tracking and budget limits
---
## ๐๏ธ Architecture
```text
claw-code/
โโโ README.md
โโโ pyproject.toml
โโโ .gitignore
โโโ images/
โ โโโ logo.png
โโโ src/ # Python implementation
โ โโโ main.py # CLI entry point & argument parsing
โ โโโ agent_runtime.py # Core agent loop (LocalCodingAgent)
โ โโโ agent_tools.py # Tool definitions & execution engine
โ โโโ agent_prompting.py # System prompt assembly
โ โโโ agent_context.py # Context building & CLAUDE.md discovery
โ โโโ agent_context_usage.py # Context usage estimation & reporting
โ โโโ agent_session.py # Session state management
โ โโโ agent_slash_commands.py # Local slash command processing
โ โโโ agent_types.py # Shared dataclasses & type definitions
โ โโโ openai_compat.py # OpenAI-compatible API client
โ โโโ session_store.py # Session serialization & persistence
โ โโโ permissions.py # Tool permission filtering
โ โโโ tools.py # Mirrored tool inventory
โ โโโ commands.py # Mirrored command inventory
โ โโโ ... # 75+ modules across 30+ packages
โ โโโ plugins/ # Plugin subsystem (WIP)
โ โโโ hooks/ # Hook system (WIP)
โ โโโ remote/ # Remote runtime modes (WIP)
โ โโโ voice/ # Voice mode (WIP)
โ โโโ vim/ # VIM mode (WIP)
โโโ tests/ # Unit tests
โโโ test_agent_runtime.py
โโโ test_agent_context.py
โโโ test_agent_context_usage.py
โโโ test_agent_prompting.py
โโโ test_agent_slash_commands.py
โโโ test_porting_workspace.py
```
---
## ๐ฆ Requirements
| Requirement | Details |
|-------------|---------|
| ๐ Python | `3.10` or higher |
| ๐ฅ๏ธ Model Server | `vLLM`, `Ollama`, `LiteLLM Proxy`, or `OpenRouter`, with tool calling support |
| ๐ง Model | [`Qwen/Qwen3-Coder-30B-A3B-Instruct`](https://huggingface.co/Qwen/Qwen3-Coder-30B-A3B-Instruct) (recommended) |
---
## ๐ Quick Start
### 1. Start vLLM with Qwen3-Coder
vLLM must be started with automatic tool choice enabled. Use the `qwen3_xml` parser for Qwen3-Coder tool calling:
```bash
python -m vllm.entrypoints.openai.api_server \
--model Qwen/Qwen3-Coder-30B-A3B-Instruct \
--host 127.0.0.1 \
--port 8000 \
--enable-auto-tool-choice \
--tool-call-parser qwen3_xml
```
Verify the server is running:
```bash
curl http://127.0.0.1:8000/v1/models
```
> ๐ **References:** [vLLM Tool Calling Docs](https://docs.vllm.ai/en/v0.13.0/features/tool_calling/) ยท [OpenAI-Compatible Server](https://docs.vllm.ai/en/v0.13.0/serving/openai_compatible_server.html)
### Optional: Use Ollama Instead of vLLM
`claw-code-agent` can also work with Ollama because the runtime targets an OpenAI-compatible API. Use a model that supports tool calling well.
Example:
```bash
ollama serve
ollama pull qwen3
```
Then configure:
```bash
export OPENAI_BASE_URL=http://127.0.0.1:11434/v1
export OPENAI_API_KEY=ollama
export OPENAI_MODEL=qwen3
```
Notes:
- prefer tool-capable models such as `qwen3`
- plain chat-only models are not enough for full agent behavior
- Ollama does not use the `vLLM` parser flags shown above
> ๐ **References:** [Ollama OpenAI Compatibility](https://docs.ollama.com/api/openai-compatibility) ยท [Ollama Tool Calling](https://docs.ollama.com/capabilities/tool-calling)
### Optional: Use LiteLLM Proxy
`claw-code-agent` can also work through LiteLLM Proxy because the runtime targets an OpenAI-compatible chat completions API. The routed model still needs to support tool calling for full agent behavior.
Quick start example:
```bash
pip install 'litellm[proxy]'
litellm --model ollama/qwen3
```
LiteLLM Proxy runs on port `4000` by default. Then configure:
```bash
export OPENAI_BASE_URL=http://127.0.0.1:4000
export OPENAI_API_KEY=anything
export OPENAI_MODEL=ollama/qwen3
```
Notes:
- LiteLLM Proxy gives you an OpenAI-style gateway in front of many providers
- tool use still depends on the underlying routed model and provider behavior
- if you configure a LiteLLM master key, use that instead of `anything`
> ๐ **References:** [LiteLLM Docs](https://docs.litellm.ai/) ยท [LiteLLM Proxy Quick Start](https://docs.litellm.ai/)
### Optional: Use OpenRouter
`claw-code-agent` can also work with [OpenRouter](https://openrouter.ai/), a cloud API gateway that provides access to models from OpenAI, Anthropic, Google, Meta, and others through a single OpenAI-compatible endpoint. No local model server required.
Configure:
```bash
export OPENAI_BASE_URL=https://openrouter.ai/api/v1
export OPENAI_API_KEY=sk-or-v1-your-key-here
export OPENAI_MODEL=openai/gpt-4o-mini
```
Notes:
- sign up at [openrouter.ai](https://openrouter.ai/) and create an API key under [Keys](https://openrouter.ai/keys)
- model names use the `provider/model` format (e.g. `anthropic/claude-sonnet-4`, `openai/gpt-4o`, `google/gemini-2.5-pro`)
- tool calling support varies by model โ check the [model list](https://openrouter.ai/models) for capabilities
- this sends your conversation (including file contents and shell output) to OpenRouter and the upstream provider โ do not use with repos containing secrets or sensitive data
> ๐ **References:** [OpenRouter Docs](https://openrouter.ai/docs) ยท [Supported Models](https://openrouter.ai/models) ยท [API Keys](https://openrouter.ai/keys)
### 2. Configure Environment
```bash
export OPENAI_BASE_URL=http://127.0.0.1:8000/v1
export OPENAI_API_KEY=local-token
export OPENAI_MODEL=Qwen/Qwen3-Coder-30B-A3B-Instruct
```
### Use Another Model With vLLM
If you want to try another model, keep the same `vLLM` server setup and change the `--model` value when you launch `vLLM`.
Example:
```bash
python -m vllm.entrypoints.openai.api_server \
--model your-model-name \
--host 127.0.0.1 \
--port 8000 \
--enable-auto-tool-choice \
--tool-call-parser your_parser
```
Then update:
```bash
export OPENAI_MODEL=your-model-name
```
Notes:
- the documented path in this repository is `vLLM`
- the model must support tool calling well enough for agent use
- some model families require a different `--tool-call-parser`
- slash commands such as `/help`, `/context`, and `/tools` are local and do not require the model server
### 3. Run the Agent
```bash
# Read-only question
python3 -m src.main agent \
"Read src/agent_runtime.py and summarize how the loop works." \
--cwd .
# Write-enabled task
python3 -m src.main agent \
"Create TEST_QWEN_AGENT.md with one line: test ok" \
--cwd . --allow-write
# Shell-enabled task
python3 -m src.main agent \
"Run pwd and ls src, then summarize the result." \
--cwd . --allow-shell
```
---
## ๐ ๏ธ Usage
### Agent Commands
| Command | Description |
|---------|-------------|
| `agent ` | Run the agent with a prompt |
| `agent-prompt` | Show the assembled system prompt |
| `agent-context` | Show estimated context usage |
| `agent-context-raw` | Show the raw context snapshot |
| `agent-resume ` | Resume a saved session |
### CLI Flags
| Flag | Description |
|------|-------------|
| `--cwd ` | Set the workspace directory |
| `--model ` | Override the model name |
| `--base-url ` | Override the API base URL |
| `--allow-write` | Allow the agent to modify files |
| `--allow-shell` | Allow the agent to execute shell commands |
| `--unsafe` | Allow destructive shell operations |
| `--show-transcript` | Print the full message transcript |
| `--system-prompt ` | Set a custom system prompt |
| `--append-system-prompt ` | Append to the system prompt |
| `--add-dir ` | Add extra directories to context |
### Slash Commands
These are handled **locally** before the model loop:
| Command | Aliases | Description |
|---------|---------|-------------|
| `/help` | `/commands` | Show built-in slash commands |
| `/context` | `/usage` | Show estimated session context usage |
| `/context-raw` | `/env` | Show raw environment & context snapshot |
| `/prompt` | `/system-prompt` | Render the effective system prompt |
| `/permissions` | โ | Show active tool permission mode |
| `/model` | โ | Show or update the active model |
| `/tools` | โ | List registered tools with permission status |
| `/memory` | โ | Show loaded CLAUDE.md memory bundle |
| `/status` | `/session` | Show runtime/session status summary |
| `/clear` | โ | Clear ephemeral runtime state |
```bash
python3 -m src.main agent "/help"
python3 -m src.main agent "/context" --cwd .
python3 -m src.main agent "/tools" --cwd .
python3 -m src.main agent "/status" --cwd .
```
### Utility Commands
```bash
python3 -m src.main summary # Workspace summary
python3 -m src.main manifest # Workspace manifest
python3 -m src.main commands --limit 10 # Command inventory
python3 -m src.main tools --limit 10 # Tool inventory
```
---
## ๐ง Built-in Tools
The agent has access to 7 core tools:
| Tool | Description | Permission |
|------|-------------|------------|
| `list_dir` | List files and directories | ๐ข Always |
| `read_file` | Read file contents (with line ranges) | ๐ข Always |
| `write_file` | Write or create files | ๐ก `--allow-write` |
| `edit_file` | Edit files via exact string matching | ๐ก `--allow-write` |
| `glob_search` | Find files by glob pattern | ๐ข Always |
| `grep_search` | Search file contents by regex | ๐ข Always |
| `bash` | Execute shell commands | ๐ด `--allow-shell` |
---
## ๐ Session Persistence
Each `agent` run automatically saves a resumable session:
```text
session_id=4f2c8c6f9c0e4d7c9c7b1b2a3d4e5f67
session_path=.port_sessions/agent/4f2c8c6f...
```
Resume a previous session:
```bash
python3 -m src.main agent-resume \
4f2c8c6f9c0e4d7c9c7b1b2a3d4e5f67 \
"Continue the previous task and finish the missing parts."
```
Inspect saved sessions:
```bash
ls -lt .port_sessions/agent
```
> **Note:** Run `agent-resume` from the same `claw-code/` directory where the session was created. A resumed session continues from the saved transcript, not from scratch.
---
## ๐งช Testing
Run the full test suite:
```bash
python3 -m unittest discover -s tests -v
```
Smoke tests:
```bash
python3 -m src.main agent "/help"
python3 -m src.main agent-context --cwd .
python3 -m src.main agent \
"Read src/agent_session.py and summarize the message flow." \
--cwd .
```
---
## ๐ Permission Model
Claw Code Agent uses a **tiered permission system** to keep the agent safe by default:
| Tier | Capability | Flag Required |
|------|-----------|---------------|
| **Read-only** | List, read, glob, grep | None (default) |
| **Write** | + file creation and editing | `--allow-write` |
| **Shell** | + shell command execution | `--allow-shell` |
| **Unsafe** | + destructive shell operations | `--unsafe` |
## ๐ Detailed Parity Status Against npm `src`
The full implementation checklist now lives in [PARITY_CHECKLIST.md](PARITY_CHECKLIST.md).
It breaks parity down by:
- core agent runtime
- CLI/runtime modes
- prompt assembly
- context and memory
- slash commands
- built-in tools
- commands and task systems
- permissions, hooks, and policy
- MCP, plugins, and skills
- interactive REPL / TUI
- remote, background, and team features
- editor, platform, and native integrations
- services and internal subsystems
- mirrored workspace versus working runtime
- high-priority next steps
---
## โ ๏ธ Disclaimer
- This repository is a **Python reimplementation** inspired by the Claude Code npm architecture.
- It does **not** ship the original npm source.
- It is **not** affiliated with or endorsed by Anthropic.
---
Built with ๐ Python ยท Powered by ๐ HarnessLab Team.