Supported Frameworks

LangChain (Python) with aim-core

from aim_sdk import secure, AgentType
from langchain.agents import AgentExecutor

# AIM auto-detects LangChain from your imports
agent = secure("research-agent", agent_type=AgentType.LANGCHAIN)

@agent.perform_action(capability="web:search", resource="search-api")
def search_web(query: str):
    """AIM verifies capability before execution"""
    return search_tool.run(query)

# All actions are logged to the tamper-evident audit trail
executor = AgentExecutor(agent=llm_agent, tools=[search_web])

CrewAI with Multi-Agent Monitoring

from aim_sdk import secure, AgentType
from crewai import Agent, Crew, Task

# Each crew member gets its own identity and trust score
researcher = secure("researcher", agent_type=AgentType.CREWAI)
writer = secure("writer", agent_type=AgentType.CREWAI)

# AIM tracks inter-agent communication and capability usage
crew = Crew(
    agents=[researcher_agent, writer_agent],
    tasks=[research_task, write_task]
)
# Crew-level trust score aggregates individual agent scores

Node.js / TypeScript with @opena2a/aim-core

import { createIdentity, signPayload, verifySignature } from '@opena2a/aim-core';

// Create a new agent identity (Ed25519 key pair)
const identity = await createIdentity({ name: 'my-agent' });

// Sign an action payload
const payload = { action: 'db:read', resource: 'users', timestamp: Date.now() };
const signature = await signPayload(payload, identity.privateKey);

// Verify the signature (on the receiving end)
const valid = await verifySignature(payload, signature, identity.publicKey);
console.log('Verified:', valid); // true

Custom Framework Integration

For frameworks not listed above, use the aim-core library directly. The core library provides identity management, policy evaluation, trust scoring, and audit logging without any framework coupling. Wrap your agent's action methods with AIM's capability checks to enforce access control regardless of the underlying framework.

from aim_sdk import secure

# Works with any Python agent framework
agent = secure("custom-agent")

# Manually check capabilities before executing actions
if agent.has_capability("file:write"):
    write_to_file(data)
else:
    agent.request_capability("file:write", reason="Need to save results")