ai-agent-governance

Python and TypeScript SDKs for verifiable evidence of AI agent actions. Signed receipts, policy enforcement, audit trails. Works with LangChain, CrewAI, MCP.

Deterministic pre-action authorization for AI agents. Guardrails run before tool execution, so prompt injection cannot bypass policy checks.

CI/CD compliance scanner for AI agents. EU AI Act, DORA, ISO 42001. GitHub Action for automated compliance checks on every PR.

Agentic OS Control Plane: deterministic state-machine governance for AI coding agents with cryptographically enforced SSH/CIBA human-in-the-loop gates and zero-trust verification.

Tool-agnostic governance framework for AI-assisted software development. Shared project rules, human approvals, structured reviews and documented enforcement capabilities — adaptable through AI client packs and project overlays.

Local-first AI agent governance with verifiable, tamper-evident audit trails. Wrap any agent CLI, anchor the journal at an RFC 3161 authority, export evidence anyone can verify. Zero dependencies.

Portfolio of Naitik Gupta, a KIET CSE student building React/Next.js apps, AI-agent governance systems, and Python machine-learning projects. Explore selected projects, skills, achievements, and certifications.

The control plane for AI agents: identity, policy, budgets, human approval and signed receipts for every agent, plus an MCP gateway, RAG chatbots, AI Act and ISO 42001 evidence, a security console and a CMS. One Go binary, no dependencies.

Coding Agent Governance: deterministic scope, evidence, and closure for AI coding-agent workflows.

UnitCause is a real-time cost governance platform for AI agents. It provides budget enforcement, loop detection, cost analytics, and policy management via a web dashboard, REST API, and Python SDK. Set budgets, stop runaway agents, and monitor spending with ease.

AI agent runtime governance control plane: intercept tool calls with PII protection, approvals, and formal verification.

Bounded authority handoff between humans and AI-assisted systems.

Codex-native OZM skill pack for AI coding agent governance, agentic coding loops, claim ceilings, and AGENTS.md-aware workflows.

Pluggable deterministic gate engine — neurosymbolic pre-action authorization: one domain-agnostic core, N domain-specific precondition gates (lease handover, cross-border data transfer, medical dispensing, AI investment).

Reference architecture: 9 governed AI agents on an 8 GB Mac mini, 99/99-suite release gate, 1,026 tests. Sanitised showcase of Empire OS.

Unit27 public protocol package for Authority-Reach-Delegation classification before AI agents receive tools, credentials, autonomy, or delegation rights.

Apache-2.0 offline verifier for AAES evidence exports (aaes.export/v2). Check hashes, signatures, and anchors without contacting AAES.

Local-first desktop app for AI agent governance and audit. Observe the timeline, inspect evidence bundles and RFC 3161 anchors, watch wrapped agent sessions, manage policy. Companion to the soma runtime.

Mandate Layer: trading mandates for AI agents, enforced by a contract on Base (Colosseum Crypto World's Fair, Base track)

A deterministic synthetic enterprise for testing AI agents against changing business context.