Invoice Match Custom
blueprint · 4 agents · 3 MCP servers
1145★ stars
149forks
4.7rating
What it does
An extractor reads the invoice document and pulls vendor, line items, quantities, and totals. A matcher checks those against the purchase order and goods-receipt records in the database: does the price match the PO, was the quantity received, has this invoice number been paid before. The flow branches on the result: a clean match is summarized and paused for human approval; a mismatch produces an exception report that names each disagreement. No payment is executed by the workflow under any path.
The cast · 4 agents
Matchermatcherreasoning
Three-way match the invoice against the purchase order and goods-receipt records: price per line against the PO, quantity against what was received, and the invoice number against payment history for duplicates. Return an object with a 'clean' boolean and, per line, what matched or disagreed. Read-only.
db.query_dsl
Extractorextractorfast
Read the invoice document. Extract the vendor, invoice number, date, line items with quantities and unit prices, and the total. Flag anything illegible rather than guessing.
drive.read_doc
Approval ClerkapprovalClerkfast
Summarize the clean match in three lines — vendor, amount, what it matched against — and post it to the AP channel for sign-off. Do not pay anything.
chat.post_message
Exception WriterexceptionWriterreasoning
Write the exception: name each disagreement precisely (duplicate invoice number, price above PO, quantity short of receipt), what the records say, and what to check first. Post it to the AP channel.
chat.post_message
Flow
→ Extractor → invoice
→ Matcher → match
⑂ branch
match.clean == true
→ Approval Clerk → disposition
→ ask → disposition
Interface
Inputs
invoiceRefstring
The invoice to process, as a drive document id or path.
Outputs
dispositionstring
Approved-and-queued, or the exception report, whichever path ran.
Depends on · 3 MCP servers
Multi-account MCP for Gmail, Calendar, Drive, Docs, and Sheets — 61 tools, tokenio.github.adelaidasofia/google-workspace-mcpdriveAVerified
io.github.supabase/postgres-mcpnot in the registry yetdb
io.github.slack/slack-mcpnot in the registry yetchat⤓ Export runnable code
# Generated by @socketcat/compiler for target: langgraph
# blueprint: com.socketcat/invoice-match 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):
invoiceRef: object
disposition: object
invoice: object
match: object
def node_AgentInvoke_0(state):
return _rt.run_agent("extractor", state, {"invoiceRef":"invoiceRef"}, "invoice")
def node_AgentInvoke_1(state):
return _rt.run_agent("matcher", state, {"invoice":"invoice"}, "match")
def node_Branch_2(state):
return {}
def node_Join_3(state):
return {}
def node_AgentInvoke_4(state):
return _rt.run_agent("approvalClerk", state, {"match":"match","invoice":"invoice"}, "disposition")
def node_Ask_5(state):
return _rt.run_ask("Approve this invoice for payment, or send it back with notes.", state, "disposition")
def node_AgentInvoke_6(state):
return _rt.run_agent("exceptionWriter", state, {"match":"match","invoice":"invoice"}, "disposition")
def route_Branch_2(state):
if _rt.cond("match.clean == true", state): return "AgentInvoke_4"
return "AgentInvoke_6"
def build():
b = StateGraph(State)
b.add_node("AgentInvoke_0", node_AgentInvoke_0)
b.add_node("AgentInvoke_1", node_AgentInvoke_1)
b.add_node("Branch_2", node_Branch_2)
b.add_node("Join_3", node_Join_3)
b.add_node("AgentInvoke_4", node_AgentInvoke_4)
b.add_node("Ask_5", node_Ask_5)
b.add_node("AgentInvoke_6", node_AgentInvoke_6)
b.add_edge(START, "AgentInvoke_0")
b.add_edge("AgentInvoke_0", "AgentInvoke_1")
b.add_edge("AgentInvoke_1", "Branch_2")
b.add_conditional_edges("Branch_2", route_Branch_2, {"AgentInvoke_4": "AgentInvoke_4", "AgentInvoke_6": "AgentInvoke_6"})
b.add_edge("Join_3", END)
b.add_edge("AgentInvoke_4", "Ask_5")
b.add_edge("Ask_5", "Join_3")
b.add_edge("AgentInvoke_6", "Join_3")
return b.compile()
INPUTS = ["invoiceRef"]
OUTPUTS = ["disposition"]
if __name__ == "__main__":
_rt.main(build, INPUTS, OUTPUTS)
▸blueprint.json (the portable format)
{
"id": "com.socketcat/invoice-match",
"flow": {
"type": "sequence",
"blocks": [
{
"in": {
"invoiceRef": "invoiceRef"
},
"out": "invoice",
"use": "extractor",
"type": "agent"
},
{
"in": {
"invoice": "invoice"
},
"out": "match",
"use": "matcher",
"type": "agent"
},
{
"else": [
{
"in": {
"match": "match",
"invoice": "invoice"
},
"out": "disposition",
"use": "exceptionWriter",
"type": "agent"
}
],
"type": "branch",
"cases": [
{
"when": "match.clean == true",
"blocks": [
{
"in": {
"match": "match",
"invoice": "invoice"
},
"out": "disposition",
"use": "approvalClerk",
"type": "agent"
},
{
"out": "disposition",
"type": "ask",
"prompt": "Approve this invoice for payment, or send it back with notes."
}
]
}
]
}
]
},
"tags": [
"finance",
"accounts-payable",
"three-way-match",
"human-in-the-loop"
],
"title": "Invoice Match",
"agents": {
"matcher": {
"model": {
"hint": "reasoning"
},
"title": "Matcher",
"tools": [
"db.query_dsl"
],
"output": {
"type": "object",
"required": [
"clean"
],
"properties": {
"clean": {
"type": "boolean"
},
"lines": {
"type": "array"
}
}
},
"instructions": "Three-way match the invoice against the purchase order and goods-receipt records: price per line against the PO, quantity against what was received, and the invoice number against payment history for duplicates. Return an object with a 'clean' boolean and, per line, what matched or disagreed. Read-only."
},
"extractor": {
"model": {
"hint": "fast"
},
"title": "Extractor",
"tools": [
"drive.read_doc"
],
"instructions": "Read the invoice document. Extract the vendor, invoice number, date, line items with quantities and unit prices, and the total. Flag anything illegible rather than guessing."
},
"approvalClerk": {
"model": {
"hint": "fast"
},
"title": "Approval Clerk",
"tools": [
"chat.post_message"
],
"instructions": "Summarize the clean match in three lines — vendor, amount, what it matched against — and post it to the AP channel for sign-off. Do not pay anything."
},
"exceptionWriter": {
"model": {
"hint": "reasoning"
},
"title": "Exception Writer",
"tools": [
"chat.post_message"
],
"instructions": "Write the exception: name each disagreement precisely (duplicate invoice number, price above PO, quantity short of receipt), what the records say, and what to check first. Post it to the AP channel."
}
},
"estate": {
"match": {
"type": "object",
"description": "The matcher's verdict, including a 'clean' flag and per-line results."
},
"invoice": {
"type": "object",
"description": "The extracted vendor, line items, and totals."
}
},
"$schema": "socketcat.dev/blueprint/v0",
"authors": [
{
"url": "https://socketcat.com",
"name": "SocketCat"
}
],
"license": "MIT",
"servers": [
{
"ref": "io.github.adelaidasofia/google-workspace-mcp",
"alias": "drive"
},
{
"ref": "io.github.supabase/postgres-mcp",
"alias": "db"
},
{
"ref": "io.github.slack/slack-mcp",
"alias": "chat"
}
],
"summary": "Extracts an invoice, three-way matches it against PO and receipt records, and routes to approval or a written exception.",
"targets": [
"langgraph",
"*"
],
"version": "1.0.0",
"interface": {
"inputs": {
"invoiceRef": {
"type": "string",
"description": "The invoice to process, as a drive document id or path."
}
},
"outputs": {
"disposition": {
"type": "string",
"description": "Approved-and-queued, or the exception report, whichever path ran."
}
}
},
"extensions": {
"dev.langgraph": {
"checkpointer": "memory"
}
},
"description": "An extractor reads the invoice document and pulls vendor, line items, quantities, and totals. A matcher checks those against the purchase order and goods-receipt records in the database: does the price match the PO, was the quantity received, has this invoice number been paid before. The flow branches on the result: a clean match is summarized and paused for human approval; a mismatch produces an exception report that names each disagreement. No payment is executed by the workflow under any path."
}More in Finance & legal
Browse all →Contract Redline
com.socketcat/contract-redline
LangGraph
Extracts a contract's clauses, checks each against your playbook, and loops a redliner against a reviewer until it's clean — then a lawyer approves.
★ 11243 servers
⑂ 146