This page covers practical setup for using the Python DeepAgents ACP server with IDEs and development tools.
Run the ACP server from the command line:
# Using the module directly
python -m deepagents_acp --root-dir /path/to/project
# Without --root-dir defaults to current working directory
python -m deepagents_acp
| Option | Description |
|---|---|
--root-dir <path> |
Root directory accessible to the agent (default: cwd) |
| Variable | Description |
|---|---|
ANTHROPIC_API_KEY |
API key for Anthropic/Claude models (required) |
LANGSMITH_TRACING |
Set to "true" to enable LangSmith tracing |
LANGSMITH_API_KEY |
API key for LangSmith |
LANGSMITH_PROJECT |
LangSmith project name |
The package includes a run.sh entrypoint for Zed integration. Add to your Zed settings:
{
"agent_servers": {
"DeepAgents": {
"type": "custom",
"command": "/absolute/path/to/deepagents/libs/acp/run.sh"
}
}
}
Make sure the script is executable:
chmod +x run.sh
If you have uv installed, you can run the server directly:
{
"agent": {
"profiles": {
"deepagents": {
"name": "DeepAgents",
"command": "uv",
"args": ["run", "python", "-m", "deepagents_acp"],
"env": {
"ANTHROPIC_API_KEY": "sk-ant-..."
}
}
}
}
}
git clone git@github.com:langchain-ai/deepagents.git
cd deepagents/libs/acp
uv sync
.env file with your API key:ANTHROPIC_API_KEY="your-key-here"
Add the agent to your Zed settings (see above)
Open Zed's Agents Panel (CMD + Shift + ?) and create a new DeepAgent thread
| Method | Description |
|---|---|
initialize |
Negotiate versions and capabilities |
session/new |
Create a new conversation session |
session/set_mode |
Switch between operating modes |
session/prompt |
Process user prompts |
session/cancel |
Cancel ongoing operations |
| Update | Description |
|---|---|
agent_message_chunk |
Stream agent text responses |
tool_call |
Notify about tool invocations |
tool_call_update |
Update tool call status |
plan |
Send task plan entries |
| Capability | Description |
|---|---|
promptCapabilities.image |
Image content support |