Guide

Migrating to deepcrew-ai

A concept map for teams coming from CrewAI or Google's Agent Development Kit (ADK), plus what deepcrew adds once you're here.

Migration

From CrewAI

CrewAI concept	deepcrew equivalent	Notes
`Agent(role=, goal=, backstory=)`	`Agent(name=, system_prompt=)`	Fold role/goal/backstory into one system prompt.
`Crew(process=Process.sequential)`	`WorkflowBuilder().add_agent(...).then(...)`	Explicit DAG edges instead of implicit list order.
`Crew(process=Process.hierarchical)`	`Orchestrator(agents=[...])`	An LLM router picks single-agent or parallel fan-out.
`@tool` / `BaseTool`	`@tool` on a plain function	Schema auto-generated from type hints.
`output_pydantic=Model`	`response_model=Model`	Result lands on `AgentResult.parsed`; one repair attempt on invalid JSON.
`Task(human_input=True)`	`AgentHooks(approve_tool=...)`	Per-tool-call, not per-task; return `False` to deny.
`Crew(verbose=True)`	`StreamPolicy.verbose()`	Streaming events, not console printing.

Before (CrewAI)

python

from crewai import Agent, Crew, Task, Process

researcher = Agent(role="Researcher", goal="Find facts", backstory="...")
crew = Crew(agents=[researcher], tasks=[Task(description="...", agent=researcher)],
            process=Process.sequential)
result = crew.kickoff()

After (deepcrew)

python

from deepcrew import Agent, run_agent

researcher = Agent(name="researcher", model="openai/gpt-4o", system_prompt="You find facts.")
result = await run_agent(researcher, [{"role": "user", "content": "..."}])

Structured output

python

# CrewAI
researcher = Agent(role="Researcher", goal="...", output_pydantic=Report)

# deepcrew
researcher = Agent(name="researcher", model="openai/gpt-4o", response_model=Report)
result = await run_agent(researcher, messages)
result.parsed  # a validated Report instance

Migration

From Google ADK

ADK concept	deepcrew equivalent	Notes
`LlmAgent(model=, instruction=)`	`Agent(model=, system_prompt=)`	`model` is a LiteLLM string, e.g. `"openai/gpt-4o"`.
ADK `FunctionTool`	`@tool` function or `Skill`	Simple callables become tools; multi-step capabilities become Skills.
ADK callbacks	`AgentHooks` + `StreamEvent` queue	Hooks intercept (can deny); events only observe.
ADK `Session` / state	`MemoryProvider`	InMemory, File, or Redis-backed.
ADK `SequentialAgent`/`ParallelAgent`	`WorkflowBuilder`	Independent DAG nodes run in parallel automatically.
ADK `LoopAgent`	`LoopConfig` + `run_agent_loop`	Verifier-driven convergence rather than a fixed iteration count.

Before (ADK)

python

from google.adk.agents import LlmAgent

agent = LlmAgent(model="gemini-2.0-flash", name="assistant", instruction="You are helpful.")

After (deepcrew)

python

from deepcrew import Agent, run_agent

agent = Agent(name="assistant", model="gemini/gemini-2.0-flash", system_prompt="You are helpful.")
result = await run_agent(agent, [{"role": "user", "content": "Hello!"}])

Migration

What deepcrew adds

True token streaming with selectable visibility — every agent, tool call, memory op, retry, and verifier score is a StreamEvent. StreamPolicy controls what a given UI sees without changing execution.
Self-improving loop — a Verifier critiques each iteration and drives refinement, with adaptive early-stopping and self-consistency branching.
Bounded recursive spawning — agents dynamically spawn sub-agents mid-run via a spawn_agent meta-tool, capped by a hard max_spawn_depth.
Skill distillation — a converged, high-confidence loop result can be distilled into a replayable Skill, Voyager-style.
Multimodal input — image()/pdf()/user_message() attach images and documents as standard content blocks.

See the Features index for the full list of what deepcrew-ai can do.