Framework Adapter Comparison¶

Edictum ships seven framework adapters. This guide helps you choose the right one and understand the tradeoffs.

When to use this¶

Read this page if you are choosing between agent frameworks, migrating from one to another, or need to understand which postcondition effects each adapter supports. If you already know your framework, go directly to its adapter page. The comparison table and per-adapter notes below cover the capability differences that matter most: whether an adapter can redact tool results before the LLM sees them, how denial is communicated, and which postcondition effects are supported.

Quick Comparison¶

Framework	Integration Method	Can Redact Before LLM	Deny Mechanism	Cost (same task)
LangChain	`as_tool_wrapper()`	Yes	Return "DENIED: reason" as ToolMessage	$0.025
OpenAI Agents	`as_guardrails()`	Deny only (`reject_content`)	`reject_content(reason)`	$0.018
CrewAI	`register()`	No (side-effect only)	before_hook returns "DENIED: reason"	$0.040
Agno	`as_tool_hook()`	Yes (hook wraps execution)	Hook returns denial string	N/A
Semantic Kernel	`register(kernel)`	Yes (filter modifies FunctionResult)	Filter sets terminate (configurable) + error	$0.008
Claude SDK	`to_hook_callables()`	No (side-effect only)	Returns deny dict to SDK	N/A
Nanobot	`wrap_registry()`	Yes (wraps execute)	Returns `"[DENIED] reason"` string	N/A

Cost column reflects benchmarks from edictum-demo using each framework's default model. N/A indicates no published benchmark data.

Which Adapter Should I Use?¶

Need full PII interception? -- Use LangChain, Agno, Semantic Kernel, or Nanobot. These adapters can replace the tool result before the LLM sees it.
Cheapest per-task cost? -- Semantic Kernel ($0.008 per task in benchmarks).
Simplest integration? -- Claude SDK or Agno. Both require minimal wiring.
Using CrewAI? -- CrewAI adapter is the only option. Note that CrewAI hooks are global (applied to every tool across all agents in the crew).

Per-Adapter Snippets¶

LangChain¶

from edictum import Edictum, Principal
from edictum.adapters.langchain import LangChainAdapter
from langgraph.prebuilt import ToolNode

guard = Edictum.from_yaml("contracts.yaml")
adapter = LangChainAdapter(guard=guard, principal=Principal(role="analyst"))
wrapper = adapter.as_tool_wrapper()
# Pass to: ToolNode(tools=tools, wrap_tool_call=wrapper)

OpenAI Agents¶

from edictum import Edictum, Principal
from edictum.adapters.openai_agents import OpenAIAgentsAdapter
from agents import function_tool

guard = Edictum.from_yaml("contracts.yaml")
adapter = OpenAIAgentsAdapter(guard=guard, principal=Principal(role="assistant"))
input_gr, output_gr = adapter.as_guardrails()
# Pass to: @function_tool(tool_input_guardrails=[input_gr], tool_output_guardrails=[output_gr])

CrewAI¶

from edictum import Edictum, Principal
from edictum.adapters.crewai import CrewAIAdapter

guard = Edictum.from_yaml("contracts.yaml")
adapter = CrewAIAdapter(guard=guard, principal=Principal(role="researcher"))
adapter.register()
# Hooks are now globally registered for all CrewAI tool calls

Agno¶

from edictum import Edictum, Principal
from edictum.adapters.agno import AgnoAdapter

guard = Edictum.from_yaml("contracts.yaml")
adapter = AgnoAdapter(guard=guard, principal=Principal(role="assistant"))
hook = adapter.as_tool_hook()
# Pass to: Agent(tool_hooks=[hook])

Semantic Kernel¶

from edictum import Edictum, Principal
from edictum.adapters.semantic_kernel import SemanticKernelAdapter

guard = Edictum.from_yaml("contracts.yaml")
adapter = SemanticKernelAdapter(guard=guard, principal=Principal(role="analyst"))
adapter.register(kernel)
# Filter is now registered on the kernel instance

Claude SDK¶

from edictum import Edictum, Principal
from edictum.adapters.claude_agent_sdk import ClaudeAgentSDKAdapter

guard = Edictum.from_yaml("contracts.yaml")
adapter = ClaudeAgentSDKAdapter(guard=guard, principal=Principal(role="sre"))
hooks = adapter.to_hook_callables()
# Use in your agent loop — see bridge recipe in Claude SDK adapter docs

Nanobot¶

from edictum import Edictum, Principal
from edictum.adapters.nanobot import NanobotAdapter

guard = Edictum.from_yaml("contracts.yaml")
adapter = NanobotAdapter(guard=guard, principal=Principal(role="user"))
governed_registry = adapter.wrap_registry(agent.tool_registry)
# Replace: agent.tool_registry = governed_registry

Known Limitations¶