attestify-langchain

Official LangChain toolkit for Attestify — governed AI loop execution with signed receipts, audit trails, governance-docs semantic search, x402-native payments on Base, and Attestify Trust (no-wallet agent identity + signed evidence).

Install

pip install attestify-langchain

Requires Python ≥ 3.9 and langchain-core ≥ 0.1. Trust tools additionally need cryptography — install with pip install attestify-langchain[trust]; everything else works without it.

Quick Start

import os
from attestify_langchain import AttestifyToolkit
from langchain_openai import ChatOpenAI
from langchain.agents import create_tool_calling_agent, AgentExecutor
from langchain_core.prompts import ChatPromptTemplate

# 1. Create the toolkit with your API key
toolkit = AttestifyToolkit(api_key=os.environ["ATTESTIFY_API_KEY"])
tools   = toolkit.get_tools()

# 2. Build the agent
llm    = ChatOpenAI(model="gpt-4o", temperature=0)
prompt = ChatPromptTemplate.from_messages([
    ["system", "You are an assistant with access to Attestify tools for governed execution and receipts."],
    ["placeholder", "{chat_history}"],
    ["human", "{input}"],
    ["placeholder", "{agent_scratchpad}"],
])
agent    = create_tool_calling_agent(llm, tools, prompt)
executor = AgentExecutor(agent=agent, tools=tools, verbose=True)

# 3. Run it
result = executor.invoke({"input": "How many runs have I used this month?"})
print(result["output"])

Available Tools

Getting Your API Key

1. Subscribe at attestifyos.com/pricing — or register free for Trust-only use at attestifyos.com/trust
2. Your att_live_... key arrives by email immediately after payment (or instantly for Trust's free tier)
3. Set it as an environment variable: export ATTESTIFY_API_KEY=att_live_...

Toolkit Options

toolkit = AttestifyToolkit(
    api_key="att_live_...",
    base_url="https://attestifyos.com",           # default
    timeout_s=60,                                 # default
    max_retries=2,                                # default
    include_control_tower=False,                  # set False for Builder-plan agents
    include_rag=False,                            # set False to omit the governance-docs search tool
    trust_agent_id="agt_...",                      # or set TRUST_AGENT_ID
    trust_private_key="...",                       # or set TRUST_PRIVATE_KEY
    include_trust=False,                           # set False to omit the Trust tools entirely
)

Attestify Trust — no-wallet agent identity + signed evidence

A separate concern from Router execution above: no x402, no lanes, no spend. Register once, then every agent run can sign proof of what it actually did.

from attestify_langchain import provision_trust_agent, AttestifyToolkit
import os

# ONE TIME, outside any agent run — never call this per-invocation, and
# never let the agent call it itself. A fresh identity per run breaks the
# agent's own verified-active streak and adds noise to Attestify's public
# census instead of a real, citable number.
creds = provision_trust_agent(
    api_key=os.environ["ATTESTIFY_API_KEY"],
    display_name="Research Agent",
    framework="langchain",
)
print(f"Store these — TRUST_AGENT_ID={creds['agent_id']}  TRUST_PRIVATE_KEY={creds['private_key']}")

# From then on, with those two env vars set:
toolkit = AttestifyToolkit(api_key=os.environ["ATTESTIFY_API_KEY"])
tools   = toolkit.get_tools()  # includes attestify_trust_submit_evidence and
                                # attestify_trust_verify automatically

The private key is bound once at AttestifyToolkit construction — it's never a tool parameter, so it never enters the LLM's own context or a transcript.

Callback Handler

Log loop completions (loop_id, cost, verified) automatically:

from attestify_langchain import AttestifyCallbackHandler

handler  = AttestifyCallbackHandler(
    on_run_complete=lambda loop_id, cost, verified:
        print(f"Loop {loop_id} cost {cost} USDC verified={verified}")
)
executor = AgentExecutor(agent=agent, tools=tools, callbacks=[handler])

LangGraph Example

from langgraph.prebuilt import create_react_agent
from langchain_openai import ChatOpenAI
from attestify_langchain import AttestifyToolkit
import os

toolkit = AttestifyToolkit(api_key=os.environ["ATTESTIFY_API_KEY"])
agent   = create_react_agent(
    ChatOpenAI(model="gpt-4o"),
    toolkit.get_tools(),
)

for chunk in agent.stream({"messages": [("human", "Submit a task: summarise Q2 revenue")]}):
    print(chunk)

(For a dedicated LangGraph package with its own state-graph node, see attestify-langgraph instead.)

Plan Gating

All plan enforcement is server-side. If an agent calls attestify_get_control_tower on a Builder plan, it receives:

{"error": "Enterprise plan required to access Control Tower.", "status": 403}

The agent can relay this directly to the user without any client-side logic.

Links

• Attestify OS
• Attestify Trust
• Pricing
• Dashboard
• Python SDK (attestify-sdk)
• GitHub