Skip to content

INTEGRATIONS.md

Latest commit

 

History

History
147 lines (102 loc) · 3.34 KB

INTEGRATIONS.md

File metadata and controls

147 lines (102 loc) · 3.34 KB

Framework Integrations

AVP works with any Python code. These integrations provide native tool adapters for popular agent frameworks.

CrewAI

pip install agentveil crewai
from agentveil.tools.crewai import AVPReputationTool, AVPDelegationTool, AVPAttestationTool

agent = Agent(
    role="Researcher",
    tools=[AVPReputationTool(), AVPDelegationTool(), AVPAttestationTool()],
)

Full example: examples/crewai_example.py

LangGraph

pip install agentveil langchain-core langgraph
from agentveil.tools.langgraph import avp_check_reputation, avp_should_delegate, avp_log_interaction
from langgraph.prebuilt import ToolNode

tool_node = ToolNode([avp_check_reputation, avp_should_delegate, avp_log_interaction])

Full example: examples/langgraph_example.py

AutoGen

pip install agentveil autogen-core
from agentveil.tools.autogen import avp_reputation_tools

agent = AssistantAgent(name="researcher", tools=avp_reputation_tools())

Full example: examples/autogen_example.py

Claude (MCP Server)

agentveil-mcp exposes Runtime Gate, approval, signed receipt, reputation, identity lookup, attestation, and audit workflows to MCP clients. Hosted read-only deployments expose public inspection tools only.

pip install 'agentveil[mcp]'

Add to claude_desktop_config.json:

{
  "mcpServers": {
    "agentveil": {
      "command": "agentveil-mcp"
    }
  }
}

Full example: examples/claude_mcp_example.py

OpenAI

pip install agentveil openai
from agentveil.tools.openai import avp_tool_definitions, handle_avp_tool_call

response = client.chat.completions.create(
    model="gpt-4", messages=messages, tools=avp_tool_definitions()
)
result = handle_avp_tool_call(tool_call.function.name, args)

Full example: examples/openai_example.py

Paperclip

pip install agentveil
from agentveil.tools.paperclip import configure, avp_should_delegate, avp_evaluate_team

configure(base_url="https://agentveil.dev", agent_name="paperclip_ceo")
avp_should_delegate(did="did:key:z6Mk...", min_score=0.5)
avp_evaluate_team(dids=["did:key:z6Mk1...", "did:key:z6Mk2..."])

Full example: examples/paperclip_example.py

Hermes (Nous Research)

AVP integrates with Hermes Agent via MCP + agentskills.io skill.

Option 1: MCP server

{
  "mcpServers": {
    "avp": {
      "command": "agentveil-mcp",
      "env": { "AVP_BASE_URL": "https://agentveil.dev" }
    }
  }
}

Option 2: Skill

cp -r skills/avp-trust-enforcement ~/.hermes/skills/

Skill file: skills/avp-trust-enforcement/SKILL.md

Any Python

No extra dependencies — use @avp_tracked decorator or AVPAgent directly.

from agentveil import avp_tracked

@avp_tracked("https://agentveil.dev", name="my_agent", to_did="did:key:z6Mk...")
def my_function(data):
    return result

Compatibility

AVP works alongside any identity provider — OAuth, API keys, custom identity solutions. Same DID standard, complementary trust layers.