Skip to main content

amcp: Lego-style coding agent CLI. Core: rg grep/find and read; editing & web search via MCP.

Project description

AMCP

PyPI version CI Python 3.11+ License: Apache-2.0

A Lego-style coding agent CLI with built-in tools (grep, read files, bash execution) and MCP server integration for extended capabilities (web search, etc.).

X: https://x.com/zhangjintao9020/status/1995170132973466018?s=20 | Demo: https://drive.google.com/file/d/1FGoY4I_JFQ1FSz19XlVJZ6Z4lWUucD7a/view?usp=sharing

Features

  • Built-in Tools: read_file, grep, bash, think, todo, apply_patch, write_file
  • MCP Integration: Connect to any MCP server for extended capabilities
  • Conversation History: Persistent sessions across runs
  • Flexible Configuration: YAML-based agent specifications
  • Tool Calling: Automatic tool selection and execution
  • ACP Support: Full Agent Client Protocol support for IDE integration (Zed, etc.)
  • AGENTS.md Support: Auto-load project-specific rules from AGENTS.md files
  • Smart Context Compaction: Intelligent context management with dynamic thresholds
  • Multi-Agent System: Primary/Subagent architecture with built-in agent types (coder, explorer, planner)
  • Event Bus: Publish/subscribe system for agent communication and extensibility
  • Hooks System: Extensible hooks for tool validation, logging, and custom behaviors

Installation

Quick Run with uvx (no install needed)

# Initialize config first (API keys, model settings)
uvx amcp-agent init

# Run the agent
uvx amcp-agent

# Run as ACP server (for IDE integration)
uvx amcp-agent acp serve

From PyPI

# Install from PyPI
pip install amcp-agent

# Or with uv (requires virtual environment or --system flag)
uv pip install amcp-agent              # in a virtual environment
uv pip install --system amcp-agent     # global install without venv

# With Anthropic Claude support
pip install amcp-agent[anthropic]

From Source (development)

# Clone the repository
git clone https://github.com/tao12345666333/amcp.git
cd amcp

# using uv (recommended)
uv venv && source .venv/bin/activate
uv pip install -e .

# or with pip
python -m venv .venv && source .venv/bin/activate
pip install -e .

Usage

# Initialize config
amcp init

# Agent with tool calling (default command)
amcp  # interactive mode with conversation history
amcp --once "create a hello.py file with a hello function"  # single message
amcp --list  # list available agent specifications
amcp --agent path/to/agent.yaml  # use custom agent spec
amcp --session my-session  # use specific session ID
amcp --clear  # clear conversation history

# MCP server management
amcp mcp tools --server exa
amcp mcp call --server exa --tool web_search_exa --args '{"query":"rust async"}'

# Run as ACP agent (for IDE integration)
amcp acp serve

ACP (Agent Client Protocol) Support

AMCP fully supports the Agent Client Protocol for integration with IDEs like Zed.

Features

  • Session Management: Create, load, and list sessions
  • Session Modes: Switch between ask, architect, and code modes
    • ask: Request permission before making changes
    • architect: Design and plan without implementation
    • code: Full tool access for implementation
  • Slash Commands: /clear, /plan, /search, /help
  • Agent Plans: Visual execution plans for complex tasks
  • Permission Requests: User approval for sensitive operations
  • Client Capabilities: Use client's filesystem and terminal when available

Running as ACP Agent

# Start the ACP agent server (stdio transport)
amcp acp serve

Zed Integration

Add to your Zed settings (~/.config/zed/settings.json):

{
  "agent": {
    "profiles": {
      "amcp": {
        "name": "AMCP",
        "provider": {
          "type": "acp",
          "command": "amcp",
          "args": ["acp", "serve"]
        }
      }
    },
    "default_profile": "amcp"
  }
}

Built-in Tools

  • read_file: Read text files with two modes:
    • Slice mode (default): Read specific line ranges
    • Indentation mode: Intelligently read code blocks around an anchor line, automatically capturing surrounding context (functions, classes)
  • grep: Search for patterns in files using ripgrep
  • bash: Execute bash commands for file operations and system tasks
  • think: Internal reasoning and planning
  • todo: Manage a todo list to track tasks during complex operations
  • apply_patch: Apply diff-based patches to files (recommended for edits, see docs/apply-patch.md)
  • write_file: Write content to files (for creating new small files)
  • task: Spawn sub-agents for parallel task execution

Config

The CLI loads MCP servers from ~/.config/amcp/config.toml. Generate a starter config:

amcp init

Example (OpenAI-compatible API):

[servers.exa]
url = "https://mcp.exa.ai/mcp"

[servers.custom]
command = "npx"
args = ["-y", "@some/mcp-server"]
env.API_KEY = "your-key"

[chat]
api_type = "openai"  # "openai" (default) or "anthropic"
base_url = "https://api.openai.com/v1"
model = "gpt-4o"
api_key = "your-api-key"
mcp_tools_enabled = true
write_tool_enabled = true  # Enable/disable built-in write_file tool

Example (OpenAI Responses API):

[chat]
api_type = "openai_responses"
model = "gpt-4o"
api_key = "your-api-key"

Example (Anthropic Claude):

[chat]
api_type = "anthropic"
model = "claude-sonnet-4-20250514"
api_key = "your-anthropic-api-key"  # or set ANTHROPIC_API_KEY env var

To use Anthropic, install with: pip install amcp-agent[anthropic]

Hooks System

AMCP provides a flexible hooks system to extend and customize agent behavior. Hooks can:

  • Validate and modify tool inputs before execution
  • Process tool outputs after execution
  • Block dangerous operations
  • Log and audit agent activities

Quick Example

Create .amcp/hooks.toml in your project:

[hooks.PreToolUse]
[[hooks.PreToolUse.handlers]]
matcher = "write_file|apply_patch"
type = "python"
script = "./scripts/validate-writes.py"
timeout = 30

[[hooks.PostToolUse.handlers]]
matcher = "*"
type = "command"
command = "echo 'Tool executed' >> /tmp/tool_log.txt"

See docs/hooks.md for full documentation.

Development

Setup Development Environment

# Clone the repository
git clone https://github.com/tao12345666333/amcp.git
cd AMCP

# Install with development dependencies
pip install -e ".[dev]"

# Install pre-commit hooks
pre-commit install

Running Tests

# Run all tests
make test

# Run with coverage
make test-cov

# Run specific test
pytest tests/test_tools.py -v

Code Quality

# Lint code
make lint

# Format code
make format

# Type check
make type-check

See CONTRIBUTING.md for detailed development guidelines.

Notes

  • rg (ripgrep) must be installed and on PATH for the grep tool.
  • MCP servers must be installed separately and runnable (stdio transport).

License

Apache-2.0

Project details


Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Source Distribution

amcp_agent-0.7.1.tar.gz (101.4 kB view details)

Uploaded Source

Built Distribution

If you're not sure about the file name format, learn more about wheel file names.

amcp_agent-0.7.1-py3-none-any.whl (114.6 kB view details)

Uploaded Python 3

File details

Details for the file amcp_agent-0.7.1.tar.gz.

File metadata

  • Download URL: amcp_agent-0.7.1.tar.gz
  • Upload date:
  • Size: 101.4 kB
  • Tags: Source
  • Uploaded using Trusted Publishing? Yes
  • Uploaded via: twine/6.1.0 CPython/3.13.7

File hashes

Hashes for amcp_agent-0.7.1.tar.gz
Algorithm Hash digest
SHA256 f30abafb900b0600ae435f1e6b54f8b5e37b40cbdd428b6cf936fad59f52a003
MD5 c44bd6e760f681a7bb921bbf72edca05
BLAKE2b-256 efc656b5672d210906881a6fbe8d5e62827c630bd02af1a2878a554db4b533d1

See more details on using hashes here.

Provenance

The following attestation bundles were made for amcp_agent-0.7.1.tar.gz:

Publisher: release.yml on tao12345666333/amcp

Attestations: Values shown here reflect the state when the release was signed and may no longer be current.

File details

Details for the file amcp_agent-0.7.1-py3-none-any.whl.

File metadata

  • Download URL: amcp_agent-0.7.1-py3-none-any.whl
  • Upload date:
  • Size: 114.6 kB
  • Tags: Python 3
  • Uploaded using Trusted Publishing? Yes
  • Uploaded via: twine/6.1.0 CPython/3.13.7

File hashes

Hashes for amcp_agent-0.7.1-py3-none-any.whl
Algorithm Hash digest
SHA256 e5654491ed111543cec7c5d9f3d9dd770309f9e77276ee3cdab7703532a5487e
MD5 37e580fda120e8ae24fecdacdd9a4bcd
BLAKE2b-256 4362c790cdbde8892e7ef8e184c60b79caecc21e81214d4c28d636f0a5bf52f8

See more details on using hashes here.

Provenance

The following attestation bundles were made for amcp_agent-0.7.1-py3-none-any.whl:

Publisher: release.yml on tao12345666333/amcp

Attestations: Values shown here reflect the state when the release was signed and may no longer be current.

Supported by

AWS Cloud computing and Security Sponsor Datadog Monitoring Depot Continuous Integration Fastly CDN Google Download Analytics Pingdom Monitoring Sentry Error logging StatusPage Status page