wire-ai 0.4.0

Workforce Intelligence & Reasoning Engine — framework-agnostic governance layer for autonomous enterprise AI agents

WIRE — Workforce Intelligence & Reasoning Engine

Framework-agnostic governance layer for autonomous enterprise AI agents.

"Describe the work. WIRE hires the workforce."

---

The Problem

Every major agent framework ships without the five things enterprises actually need:

Gap	LangGraph	CrewAI	AutoGen
Loop protection	❌	❌	❌
Tamper-proof audit	❌	❌	❌
Hard cost ceilings	❌	❌	❌
HITL as first-class primitive	Partial	❌	Unstable
Live workforce visibility	❌	❌	❌

WIRE adds all of them — without replacing your existing framework.

---

Install

pip install wire-ai                  # core
pip install wire-ai[langgraph]       # + LangGraph adapter
pip install wire-ai[crewai]          # + CrewAI adapter  (Sprint 5)
pip install wire-ai[autogen]         # + AutoGen adapter (Sprint 5)
pip install wire-ai[all]             # everything

---

Sprint 1 — 5 lines, full governance

import wire

# Your existing LangGraph graph — unchanged
from langgraph.graph import StateGraph
graph = StateGraph(...).compile()

# Wrap with WIRE
workforce = wire.deploy(
    graph,
    backend="langgraph",
    max_iterations=30,       # LoopGuard: halt before runaway
    max_cost_usd=0.50,       # Budget: hard $0.50 ceiling
    hourly_budget_usd=0.10,  # Budget: $0.10/hour rolling
    audit_path="audit.jsonl" # AuditChain: tamper-proof log
)

# Run — same API as graph.ainvoke()
result = await workforce.ainvoke({"messages": [...]})

# Verify audit integrity
wire.AuditChain.verify("audit.jsonl")
# ✓ 12 entries · chain intact

---

What You Get

LoopGuard

Halts runaway agent loops before they exhaust your API quota. Configurable iteration and cost limits. Raises LoopBreachError with full context.

# Fires when iterations > 30 OR cost > $0.50
workforce = wire.deploy(graph, max_iterations=30, max_cost_usd=0.50)

AuditChain

Every agent action, node execution, and tool call is recorded in a tamper-proof SHA-256 hash-linked log. Verify integrity at any time.

wire.AuditChain.verify("audit.jsonl")
# Raises AuditChainError with entry index if tampered

# Replay any past run
wire replay --run-id run_20260612_abc123

Budget

Hard cost ceilings with rolling hourly and daily windows. Never get a surprise bill.

budget = wire.Budget(hourly=0.50, daily=5.00)
# BudgetBreachError fires before the ceiling is exceeded

EventBus

Typed events for every runtime moment. Subscribe handlers for alerting, logging, or custom business logic.

@workforce.on(wire.EventKind.LOOP_BREACH)
async def alert(event):
    await slack.send(f"Loop breach in {event.run_id}!")

---

CLI

wire version                          # v0.1.0
wire status                           # installed adapters
wire audit audit.jsonl                # verify chain integrity
wire replay --run-id abc123           # replay a past run

---

Sprint Roadmap

Sprint	Ships	Status
S1 — MVP	deploy() + LoopGuard + AuditChain + Budget	✅ v0.1.0
S2 — Governance	HITLGate + IdempotencyGuard + SLATracker	🔜 v0.2.0
S3 — HIRE	wire.hire("...") + 20 role templates	🔜 v0.3.0
S4 — Visibility	Live dashboard + Slack HITL + time-travel replay	🔜 v0.4.0
S5 — Multi-Framework	CrewAI + AutoGen + OpenAI adapters	🔜 v0.5.0
S6 — Enterprise	SOC-2 presets + SSO + RBAC + web dashboard	🔜 v1.0.0

---

Architecture

┌──────────────────────────────────────────────────────┐
│  HIRE LAYER     intent → role matching → workforce   │
├──────────────────────────────────────────────────────┤
│  RUNTIME LAYER  LoopGuard · HITLGate · SLATracker    │
│                 IdempotencyGuard · Budget · State     │
├──────────────────────────────────────────────────────┤
│  VISIBILITY     Dashboard · AuditChain · CostLedger  │
│                 TimeTravel · DriftDetector            │
├──────────────────────────────────────────────────────┤
│  ADAPTERS       LangGraph · CrewAI · AutoGen · OpenAI│
└──────────────────────────────────────────────────────┘

---

Why WIRE?

Built from 95 adversarially-verified research claims across 24 sources:

• LangGraph self-describes as "very low-level" — no governance primitives
• AutoGen Studio is "not meant for production" — their own README
• CrewAI has no idempotency guard — payments and emails can fire twice on retry
• No framework ships a tamper-proof audit chain, built-in workforce visibility, or HITL routing

WIRE fills every gap. Framework-agnostic. Enterprise-ready from day 1.

---

Contributing

See CONTRIBUTING.md. MIT licensed.

---

Built by Naveen Kumar Baskaran