Runtime Governance for Autonomous AI

Agents act.
Reshimu decides
if they should.

When an AI agent takes an action it shouldn't, you usually find out too late. Reshimu sits between the agent and the action, and blocks the dangerous ones before they run.

Runtime governance · Irreversible action classification · Intent persistence · Ambiguity escalation

Free during beta · LangGraph · Claude Code · AutoGen · CrewAI · Zero native deps

The Problem

Agents don't fail loudly.
They fail quietly, at scale, after the damage is done.

01 — Irreversible

Your agent called DELETE.

It already ran. There is no undo. The audit trail tells you what happened, not what to do next.

02 — Drift

Ten delegations later, the output answers a different question.

No single agent is wrong. The session has quietly forgotten what it was for.

03 — Ambiguity

"Update billing" means three things to five agents.

One of them will pick the wrong one. Probably the irreversible one. Probably this week.

Example

What running SHOR
looks like.

SHOR extracts every addressable entity from the agent's output and checks each one against the context the agent was given. Below: a partial hallucination caught before it reaches a tool call.

SHOR · GROUNDING CLASSIFIER
TypeScript
import { classify } from '@reshimu/shor'

const result = classify({
  output:  'Q3 revenue was $4.2M from 47 customers.',
  context: 'Q3 numbers: 47 customers signed up, but revenue was not disclosed.',
})
● RETURNS
{
  level:          "PARTIAL",
  score:          0.6666666666666666,
  flagForReview:  true,
  explanation:    "number '$4.2M' not found in context.",
  entities: [
    { text: "Q3",           type: "date",   found: true  },
    { text: "$4.2M",        type: "number", found: false },
    { text: "47 customers", type: "number", found: true  },
  ]
}
$ npm install @reshimu/shor
Architecture

Six levels. One direction.
Every message governed.

A vertical hierarchy. Intent flows downward through five stages, each one binding the work more tightly to the original instruction. Guardians sit at every level.

ATZMUS Root Architect · receives human intent 0 GUARDIANS
עַצְמוּת— Essence
↓ First Tzimtzum — scope definition
ADAM KADMON Strategic Mind · domain decomposition 2 GUARDIANS
אָדָם קַדְמוֹן— Primordial Human
↓ Second Tzimtzum — domain binding
ATZILUT Reasoning Council · Chokhmah · Binah · Da'at 3 GUARDIANS
אֲצִילוּת— World of Emanation
↓ Third Tzimtzum — knowledge formalization
BERIAH Design Agents · schemas · contracts 5 GUARDIANS
בְּרִיאָה— World of Creation
↓ Fourth Tzimtzum — execution specification
YETZIRAH Workflow Agents · tool calls · sequencing 8 GUARDIANS
יְצִירָה— World of Formation
↓ Fifth Tzimtzum — output binding
MALKHUT Implementation · the actual work 10 GUARDIANS
מַלְכוּת— Kingship · Manifestation
01

Intent Specification

Define what the session is for in three bullets. Every downstream agent is scored against it. Drift is caught early — before twenty messages compound into the wrong outcome.

02

Runtime Enforcement

Four guardian classifiers run on every message before delivery. Sub-5ms. No LLM calls for classification. Deterministic where it has to be.

03

Full Audit Trail

Every intercept. Every decision. Every escalation. Written to disk. Queryable. Exportable. The impression that keeps the system honest.

Runtime Classifiers

Four classifiers.
One runtime.

Scope Boundary Enforcer

Rule-based · <1ms

Detects when agents exceed their assigned domain. Catches scope creep before it propagates downstream and contaminates sibling sessions.

npm install @reshimu/aryeh
Aryeh — אַרְיֵה

Grounding & Hallucination Detector

Hybrid · semantic + rule-based

Detects ungrounded confidence and hallucinated claims. Every output must trace to its source. Confidence without basis is a flag, not a result.

npm install @reshimu/shor
Shor — שׁוֹר

Irreversible Action Classifier

Rule-based · <5ms · no LLM calls

Classifies all proposed actions GREEN / YELLOW / RED before execution. Blocks dangerous calls automatically. Human-in-loop for red. Blocked

npm install @reshimu/nesher
Nesher — נֶשֶׁר

Ambiguity Escalation System

Hybrid · semantic detection

Detects genuine gray zones. Files a structured escalation report. Halts the work until a superior agent — or a human — resolves it.

Panim Adam — פָּנִים אָדָם

All four available on npm.

Works With
LangGraph Claude Code AutoGen CrewAI OpenAI Agents

Runtime governance is not optional.
It's inevitable.

Be early.

Developer beta · Free to start · No credit card

רְשִׁימוּ
For Businesses

Bring AI to your business —
built with governance from day one.

Reshimu.ai also builds custom AI automation for businesses ready to move beyond experiments. We design, deploy, and govern the systems — so you get the leverage of AI without the risk of agents acting without intention.

Assess

Audit your workflows.

Identify the right places for AI. Two-week engagement. Deliverable: full AI roadmap and risk-ranked automation map.

Build

Custom automations, governed.

Designed and deployed by us. Always governed by Reshimu's runtime layer from day one. 4–12 week engagements. Production systems, not demos.

Scale

Embedded partnership.

We become your AI infrastructure team. Ongoing. For companies treating AI as core infrastructure, not a feature flag.

Tell us what you're trying to automate →