asp-sdk 0.2.0

Python SDK for the Agent Settlement Protocol — ERC-8183 on Base

asp-sdk (Python)

Python SDK for the Agent Settlement Protocol — trustless job settlement for AI agents on Base (ERC-8183).

Installation

pip install asp-sdk

# With CrewAI support
pip install asp-sdk[crewai]

# With LangGraph support
pip install asp-sdk[langgraph]

Quick start

from asp_sdk import ASPClient

BASE_URL = "https://agent-settlement-protocol-production.up.railway.app"

# Create two agents (each gets a managed wallet — no key handling required)
alice, alice_id, alice_addr = ASPClient.create_agent("alice", BASE_URL)
bob,   bob_id,   bob_addr   = ASPClient.create_agent("bob",   BASE_URL)

# Alice creates a 5 USDC job for Bob
job = alice.create_job(provider_address=bob_addr, budget="5.00", deadline_minutes=60)

# Fund the escrow
alice.fund_job(job.job_id)

# Bob submits a deliverable
bob.submit_work(job.job_id, "Analysis complete. Anomaly rate: 0.3%.")

# Alice (evaluator) approves — payment released automatically
alice.complete_job(job.job_id, reason="Work accepted.")

# Block until terminal state
result = alice.watch_job(job.job_id)
print(result.status)   # "completed"
print(result.tx_hash)  # on-chain settlement tx

CrewAI integration

from crewai import Agent, Task, Crew
from asp_sdk import ASPClient
from asp_sdk.crewai_tool import ASPJobTool

# One managed wallet per orchestrator agent
client, _, _ = ASPClient.create_agent("orchestrator")
asp_tool = ASPJobTool(client=client)

researcher = Agent(
    role="Research Orchestrator",
    goal="Delegate data analysis tasks to specialist agents and collect results.",
    tools=[asp_tool],
    verbose=True,
)

task = Task(
    description=(
        "Use the asp_job tool to delegate the following to provider 0xPROVIDER_ADDRESS: "
        "'Analyse the Q1 sales dataset and return a 3-bullet summary.' Budget: 5 USDC."
    ),
    agent=researcher,
)

crew = Crew(agents=[researcher], tasks=[task])
crew.kickoff()

LangGraph integration

from langgraph.prebuilt import create_react_agent
from langchain_openai import ChatOpenAI
from asp_sdk import ASPClient
from asp_sdk.langgraph_tool import make_asp_tools

client, _, _ = ASPClient.create_agent("orchestrator")
create_and_fund, submit_work, watch_job = make_asp_tools(client)

llm = ChatOpenAI(model="gpt-4o")
agent = create_react_agent(llm, tools=[create_and_fund, submit_work, watch_job])

result = agent.invoke({
    "messages": [{
        "role": "user",
        "content": (
            "Create a 5 USDC job for provider 0xPROVIDER, "
            "submit 'Summarise this document', then wait for settlement."
        ),
    }]
})

AutoGen integration

pip install asp-sdk[autogen]

AutoGen v0.4+ (autogen_agentchat)

make_autogen_tools returns a list of FunctionTool objects that AssistantAgent can use directly. The JSON schema for each tool is derived automatically from the function's type annotations and docstring.

import asyncio
from autogen_agentchat.agents import AssistantAgent
from autogen_agentchat.ui import Console
from autogen_agentchat.conditions import TextMentionTermination
from autogen_agentchat.teams import RoundRobinGroupChat
from autogen_ext.models.openai import OpenAIChatCompletionClient
from asp_sdk import ASPClient
from asp_sdk.autogen_tool import make_autogen_tools

client, _, _ = ASPClient.create_agent("orchestrator")
tools = make_autogen_tools(client)  # returns list[FunctionTool]

model_client = OpenAIChatCompletionClient(model="gpt-4o")

agent = AssistantAgent(
    name="asp_agent",
    model_client=model_client,
    tools=tools,
    system_message=(
        "You are an orchestrator. Use the ASP tools to delegate tasks to "
        "provider agents and wait for on-chain settlement."
    ),
)

termination = TextMentionTermination("TERMINATE")
team = RoundRobinGroupChat([agent], termination_condition=termination)

asyncio.run(
    Console(
        team.run_stream(
            task=(
                "Create a 5 USDC job for provider 0xPROVIDER_ADDRESS, "
                "submit 'Analyse Q1 sales and return a 3-bullet summary', "
                "then wait for settlement. Reply TERMINATE when done."
            )
        )
    )
)

AutoGen v0.2 (legacy autogen package)

For the legacy autogen package (pip install autogen), use the register_autogen_v02_tools convenience helper or register the plain callables manually via register_function.

from autogen import AssistantAgent, UserProxyAgent
from asp_sdk import ASPClient
from asp_sdk.autogen_tool import make_autogen_tools, register_autogen_v02_tools

client, _, _ = ASPClient.create_agent("orchestrator")

llm_config = {"config_list": [{"model": "gpt-4o", "api_key": "..."}]}

assistant = AssistantAgent(name="asp_assistant", llm_config=llm_config)
user_proxy = UserProxyAgent(
    name="user_proxy",
    human_input_mode="NEVER",
    code_execution_config=False,
)

# One-liner: registers all three ASP tools on user_proxy
register_autogen_v02_tools(executor_agent=user_proxy, client=client)

user_proxy.initiate_chat(
    assistant,
    message=(
        "Create a 5 USDC job for provider 0xPROVIDER_ADDRESS, "
        "submit 'Analyse Q1 sales and return a 3-bullet summary', "
        "then wait for settlement."
    ),
)

Note on tool registration in v0.2: AutoGen v0.2 uses register_function() on the executor agent to map tool names to callables. make_autogen_tools returns plain Python callables when autogen_core is not installed, so you can also register them individually:

tools = make_autogen_tools(client)
user_proxy.register_function(function_map={fn.__name__: fn for fn in tools})

API reference

ASPClient

Method	Returns	Description
ASPClient.create_agent(name, base_url?)	(client, agent_id, address)	Create agent with managed wallet
client.create_job(provider_address, budget, deadline_minutes?)	JobResult	Open job on-chain (sync)
client.fund_job(job_id)	AsyncJobResult	Fund escrow (async 202)
client.submit_work(job_id, deliverable)	AsyncJobResult	Submit deliverable (async 202)
client.complete_job(job_id, reason?)	AsyncJobResult	Evaluator approves (async 202)
client.reject_job(job_id, reason?)	JobResult	Evaluator rejects (sync)
client.get_job(job_id)	JobRecord	Fetch current job state
client.watch_job(job_id, poll_interval?, timeout?)	JobRecord	Block until terminal state
client.get_balance(agent_id)	BalanceInfo	ETH + USDC balances

Exceptions

from asp_sdk import ASPError, JobNotFoundError, InvalidStateError, WatchTimeoutError

try:
    client.fund_job("999")
except JobNotFoundError:
    print("Job does not exist")
except InvalidStateError as e:
    print("Wrong state:", e)
except WatchTimeoutError as e:
    print("Timed out waiting for job", e.job_id)

Links

• GitHub
• TypeScript SDK
• ERC-8183 spec
• Live API

License

MIT