cloudagent

Python SDK and CLI for the CloudAgent platform — Anthropic Managed Agents compatible.

Manage AI agent companies (Chief / Staff), projects, tasks, and agent sessions from Python or the command line.

Installation

pip install cloudagent

Requires Python 3.10+.

Quickstart — CLI

# Configure once
cloudagent auth login --api-key $YOUR_API_KEY

# 14 subcommand groups available
cloudagent --help

# Create a company → add a staff → send a task to the CEO
cloudagent companies create --name acme --display-name "Acme Inc"
cloudagent staffs create --company acme --name engineering \
    --display-name "Engineering" --role-display "Engineering Lead" \
    --role-category backend --default-model qwen3.6-plus
cloudagent tasks create --company acme \
    --title "Q3 plan" --brief "figure it out" --chief

# Talk to the CEO (use the session_id printed by tasks create)
cloudagent messages send <session_id> \
    --to-agent <ceo_agent_id> --kind task \
    --text "and include lost-deal reasons"

# Pull the inbox
cloudagent notifications list --unread-only
cloudagent tasks timeline <task_id>

Quickstart — Python SDK

from cloudagent import CloudAgent

with CloudAgent() as client:                            # reads CLOUDAGENT_API_KEY
    # Companies
    page = client.companies.list(limit=20)
    company = client.companies.create(name="acme", display_name="Acme Inc")

    # Tasks with chief sugar
    result = client.tasks.create(
        company="acme",
        title="Q3 plan",
        brief="figure it out",
        owner="chief",                                  # SDK resolves to CEO agent
    )
    session_id = result["session_id"]
    chief_agent_id = next(
        (s["id"] for s in client.staffs.org_chart("acme")["staffs"] if s.get("is_chief")),
        None,
    )

    # Append a follow-up message to the CEO
    client.sessions.messages.create(
        session_id,
        from_agent="user",
        to_agent=chief_agent_id,
        kind="task",
        content={"text": "and include lost-deal reasons"},
    )

    # Poll for replies
    for item in client.tasks.timeline(result["task"]["id"]).data:
        print(item.get("type"), item.get("content"))

See examples/ for full snippets including error handling.

Configuration

Migration from cloudagent-cli

cloudagent-cli (≤ 0.1.3) has been superseded by this package. Imports and CLI behaviour are preserved:

cloudagent-cli will be deprecated and yanked from PyPI after a transition period.

License

MIT — see LICENSE.