Plan and Execute LangGraph
blueprint · 4 agents · 2 MCP servers
1490★ stars
208forks
4.7rating
What it does
A planner decomposes the goal into ordered steps. An executor performs the next step using real tools and reports the result. A replanner revises the remaining plan in light of that result — removing finished steps, adding newly needed ones, or marking the goal complete — and the loop repeats. A finalizer assembles the outcome. This keeps an agent coherent over long tasks that a single prompt would lose track of. Wire the executor to whatever tools your task needs.
The cast · 4 agents
Plannerplannerreasoning
Break the goal into a short, ordered list of concrete steps. Each step should be something a single execution can complete. Do not execute anything.
no tools · reasoning only
Executorexecutorcoding
Carry out the first unfinished step in the plan using the available tools. Report what you did and what you found. Do only that step.
search.webcode.get_file
Finalizerfinalizerfast
Assemble the results of the completed steps into the final deliverable the goal asked for. Keep it clean and complete.
no tools · reasoning only
Replannerreplannerreasoning
Given the plan and the latest result, revise the remaining steps: drop what's done, add what's newly needed, reorder if reality changed. Return an object with the updated steps and a 'done' boolean, set true only when the goal is fully met.
no tools · reasoning only
Flow
→ Planner → plan
↻ loop until plan.done == true · max 6
→ Executor → result
→ Replanner → plan
→ Finalizer → result
Interface
Inputs
goalstring
The multi-step goal to accomplish.
Outputs
resultstring
The finished outcome once the plan is complete.
Depends on · 2 MCP servers
io.github.brave/brave-search-mcpnot in the registry yetsearch
io.github.github/github-mcpnot in the registry yetcode⤓ Export runnable code
# Generated by @socketcat/compiler for target: langgraph
# blueprint: com.socketcat/plan-and-execute v1.0.0 schema: socketcat.dev/blueprint/v0
# This code is yours. Edit it freely. The socketcat_runtime helper is optional and can be vendored.
from typing import TypedDict
from langgraph.graph import StateGraph, START, END
import socketcat_runtime as _rt
class State(TypedDict, total=False):
goal: object
result: object
plan: object
__loop_1: object
def node_AgentInvoke_0(state):
return _rt.run_agent("planner", state, {"goal":"goal"}, "plan")
def node_LoopHeader_1(state):
return {}
def node_LoopTick_2(state):
return {"__loop_1": state.get("__loop_1", 0) + 1}
def node_Nop_3(state):
return {}
def node_AgentInvoke_4(state):
return _rt.run_agent("executor", state, {"plan":"plan"}, "result")
def node_AgentInvoke_5(state):
return _rt.run_agent("replanner", state, {"plan":"plan","result":"result"}, "plan")
def node_AgentInvoke_6(state):
return _rt.run_agent("finalizer", state, {"goal":"goal","plan":"plan"}, "result")
def route_LoopHeader_1(state):
count = state.get("__loop_1", 0)
return "exit" if (_rt.cond("plan.done == true", state) or count >= 6) else "loop"
def build():
b = StateGraph(State)
b.add_node("AgentInvoke_0", node_AgentInvoke_0)
b.add_node("LoopHeader_1", node_LoopHeader_1)
b.add_node("LoopTick_2", node_LoopTick_2)
b.add_node("Nop_3", node_Nop_3)
b.add_node("AgentInvoke_4", node_AgentInvoke_4)
b.add_node("AgentInvoke_5", node_AgentInvoke_5)
b.add_node("AgentInvoke_6", node_AgentInvoke_6)
b.add_edge(START, "AgentInvoke_0")
b.add_edge("AgentInvoke_0", "AgentInvoke_4")
b.add_conditional_edges("LoopHeader_1", route_LoopHeader_1, {"loop": "AgentInvoke_4", "exit": "Nop_3"})
b.add_edge("LoopTick_2", "LoopHeader_1")
b.add_edge("Nop_3", "AgentInvoke_6")
b.add_edge("AgentInvoke_4", "AgentInvoke_5")
b.add_edge("AgentInvoke_5", "LoopTick_2")
b.add_edge("AgentInvoke_6", END)
return b.compile()
INPUTS = ["goal"]
OUTPUTS = ["result"]
if __name__ == "__main__":
_rt.main(build, INPUTS, OUTPUTS)
▸blueprint.json (the portable format)
{
"id": "com.socketcat/plan-and-execute",
"flow": {
"type": "sequence",
"blocks": [
{
"in": {
"goal": "goal"
},
"out": "plan",
"use": "planner",
"type": "agent"
},
{
"max": 6,
"type": "loop",
"until": "plan.done == true",
"blocks": [
{
"in": {
"plan": "plan"
},
"out": "result",
"use": "executor",
"type": "agent"
},
{
"in": {
"plan": "plan",
"result": "result"
},
"out": "plan",
"use": "replanner",
"type": "agent"
}
]
},
{
"in": {
"goal": "goal",
"plan": "plan"
},
"out": "result",
"use": "finalizer",
"type": "agent"
}
]
},
"tags": [
"planning",
"orchestration",
"long-horizon",
"plan-execute"
],
"title": "Plan and Execute",
"agents": {
"planner": {
"model": {
"hint": "reasoning"
},
"title": "Planner",
"instructions": "Break the goal into a short, ordered list of concrete steps. Each step should be something a single execution can complete. Do not execute anything."
},
"executor": {
"model": {
"hint": "coding"
},
"title": "Executor",
"tools": [
"search.web",
"code.get_file"
],
"instructions": "Carry out the first unfinished step in the plan using the available tools. Report what you did and what you found. Do only that step."
},
"finalizer": {
"model": {
"hint": "fast"
},
"title": "Finalizer",
"instructions": "Assemble the results of the completed steps into the final deliverable the goal asked for. Keep it clean and complete."
},
"replanner": {
"model": {
"hint": "reasoning"
},
"title": "Replanner",
"output": {
"type": "object",
"required": [
"done"
],
"properties": {
"done": {
"type": "boolean"
},
"steps": {
"type": "array"
}
}
},
"instructions": "Given the plan and the latest result, revise the remaining steps: drop what's done, add what's newly needed, reorder if reality changed. Return an object with the updated steps and a 'done' boolean, set true only when the goal is fully met."
}
},
"estate": {
"plan": {
"type": "object",
"description": "The current plan: remaining steps and a 'done' flag."
},
"result": {
"type": "object",
"description": "The executor's result from the most recent step."
}
},
"$schema": "socketcat.dev/blueprint/v0",
"authors": [
{
"url": "https://socketcat.com",
"name": "SocketCat"
}
],
"license": "MIT",
"servers": [
{
"ref": "io.github.brave/brave-search-mcp",
"alias": "search"
},
{
"ref": "io.github.github/github-mcp",
"alias": "code"
}
],
"summary": "Plans a multi-step task, executes one step at a time, and replans from what actually happened until the goal is met.",
"targets": [
"langgraph",
"*"
],
"version": "1.0.0",
"interface": {
"inputs": {
"goal": {
"type": "string",
"description": "The multi-step goal to accomplish."
}
},
"outputs": {
"result": {
"type": "string",
"description": "The finished outcome once the plan is complete."
}
}
},
"extensions": {
"dev.langgraph": {
"checkpointer": "memory"
}
},
"description": "A planner decomposes the goal into ordered steps. An executor performs the next step using real tools and reports the result. A replanner revises the remaining plan in light of that result — removing finished steps, adding newly needed ones, or marking the goal complete — and the loop repeats. A finalizer assembles the outcome. This keeps an agent coherent over long tasks that a single prompt would lose track of. Wire the executor to whatever tools your task needs."
}More in Core patterns
Browse all →Supervisor Team
com.socketcat/supervisor-team
LangGraph
A supervisor decides who works next — a researcher who searches or a coder who edits the repo — and assembles the result when the work is done.
★ 21402 servers
⑂ 312