How It Works

Intutic is a transparent governance layer that sits between your AI agents and the LLM providers they call. It enforces policies, detects anomalies, tracks costs, and records every decision — without adding meaningful latency.

Architecture Overview

┌──────────────┐     ┌──────────────────────────────────────────┐
│  Claude Code │     │            Intutic Control Plane         │
│  Cursor      │────▶│                                          │
│  Aider       │     │  ┌──────────┐  ┌──────┐  ┌───────────┐  │
│  Windsurf    │     │  │  Proxy   │──│ PCAS │──│  Circuit  │  │
│  Antigravity │     │  │ Gateway  │  │      │  │  Breaker  │  │
│  Codex       │     │  └────┬─────┘  └──────┘  └─────┬─────┘  │
│  OpenHands   │     │       │                        │        │
│  n8n         │     │  ┌────▼─────┐  ┌──────┐  ┌─────▼─────┐  │
└──────────────┘     │  │  FinOps  │  │  ARE │  │   SOP     │  │
                     │  │  Ledger  │  │      │  │  Registry │  │
                     │  └──────────┘  └──────┘  └───────────┘  │
                     └──────────────────────────────────────────┘
                                      │
                              ┌───────▼───────┐
                              │  LLM Provider │
                              │  (Anthropic,  │
                              │   OpenAI, …)  │
                              └───────────────┘

The Proxy Gateway

Every LLM request from your agents flows through the Intutic proxy. The proxy is transparent — agents don't need to change their code. The CLI's init command configures each harness to route through the proxy by setting the appropriate base URL or config variable.

How routing works per harness:

Harness	Config File	Mechanism
Cursor	.cursorrules	Markdown rules injected with proxy URL
Claude Code	CLAUDE.md	Markdown rules injected with proxy URL
Windsurf	.windsurfrules	Markdown rules injected with proxy URL
Aider	.aider.conf.yml	extra-instructions YAML field
Antigravity	.gemini/settings.json	customInstructions JSON field
Codex	.env.intutic	ANTHROPIC_BASE_URL / OPENAI_BASE_URL env vars
OpenHands	config.toml	[intutic] TOML section with proxy_url
n8n	(API-based)	Workflow parameters via n8n REST API (Phase 2)

Enforcement Actions (PCAS)

The Policy Compliance and Action System evaluates every request against your SOPs and applies one of four enforcement actions:

Action	What happens	When it's used
BYPASS	Request passes through unmodified	Compliant with all SOPs
ENHANCE	Request is modified (prompt enrichment, model upgrade)	SOP suggests improvements
HIJACK	Request is rerouted to a different model or modified substantially	Cost optimization, capability routing
KILL	Request is blocked entirely	Policy violation, budget breach, anomaly detected

The Circuit Breaker

The circuit breaker is the runtime enforcement mechanism. It evaluates each request against:

1. SOP rules — Does this request comply with active SOPs?
2. Budget limits — Is the user/workspace within budget tier limits?
3. Anomaly scores — Has the ARE flagged this session?
4. Trust scores — What's the trust level of this agent session?

If any check fails, the circuit breaker applies the appropriate enforcement action.

SOP Lifecycle

SOPs (Standard Operating Procedures) are the policy documents that define governance rules. They follow a 7-state lifecycle:

DRAFT → PENDING_REVIEW → GENERATED → HYPOTHESIZED → REFINED → VALIDATED
                                                                    ↓
                                                              INVALIDATED

State	Meaning
DRAFT	Initial authoring, not yet active
PENDING_REVIEW	Submitted for team review
GENERATED	Auto-generated from observed patterns
HYPOTHESIZED	Proposed rule being tested
REFINED	Iteratively improved based on feedback
VALIDATED	Active and enforced
INVALIDATED	Retired or superseded

SOPs include:

• Risk tier — LOW, MEDIUM, HIGH, CRITICAL
• Complexity tier — task complexity classification
• Dependencies — other SOPs this one depends on
• Markdown content — the actual policy rules

Changes to SOPs are classified as STRENGTHEN, CLARIFY, NARROW, or WEAKEN for audit trail.

Anomaly Detection (ARE)

The Autonomous Reasoning Engine detects 12 categories of runtime anomalies:

Anomaly Type	Description
TOOL_ABUSE	Excessive or inappropriate tool calls
TOKEN_WASTE	Inefficient token usage patterns
LOOP_DETECTED	Agent stuck in a retry/repeat loop
UNAUTHORIZED_TOOL	Calling tools outside allowed set
DATA_EXFILTRATION	Attempting to leak sensitive data
PROMPT_INJECTION	Malicious prompt manipulation detected
HALLUCINATION	Model generating fabricated information
SCOPE_VIOLATION	Operating outside defined task scope
BUDGET_BREACH	Exceeding allocated budget
SPAWN_BUDGET_BREACH	Sub-agent spawning over limits
WORKFLOW_BUDGET_BREACH	Multi-step workflow over budget
WORKFLOW_GOAL_DRIFT	Workflow deviating from stated objective

FinOps Ledger

Every execution trace records:

• Input/output token counts
• Model used and actual cost in USD
• Enforcement action applied
• Token utility classification (USEFUL or WASTED)

Budget tiers control spending limits per developer level:

Tier	Role
JUNIOR	Junior developers — lowest budget ceiling
SENIOR	Senior developers
STAFF	Staff engineers
PRINCIPAL	Principal engineers — highest budget ceiling

Sync Daemon

The intutic connect command starts a long-lived sync daemon that:

1. Polls the control plane for SOP updates (default: every 30 seconds)
2. Detects which harnesses are present in the workspace
3. Writes updated governance config to each harness's config file
4. Reports sync state back to the control plane
5. Uses atomic writes (tmp file + rename) to prevent file corruption

The daemon supports all 8 harness adapters and handles each one's config format natively.