# Agnys

> Agnys is evidence infrastructure for AI agents. It captures model calls and operational actions, connects them into traces, and provides observability, integrity checks, replay, search and evidence exports.

Canonical website: https://agnys.net/
Product application: https://app.agnys.net/
Contact: contact@agnys.net

## Core pages

- [AI agent audit trails](https://agnys.net/ai-agent-audit-trail/): How Agnys records model calls, tool use, file edits, commands, approvals and outcomes in an ordered trace.
- [AI agent observability](https://agnys.net/ai-agent-observability/): Live activity, behavioral baselines, anomaly detection, search, cost analysis and session replay.
- [EU AI Act Article 12 logging](https://agnys.net/eu-ai-act/article-12-ai-logging/): A primary-source-based explanation of record-keeping for high-risk AI systems and the operational evidence Agnys can support.
- [About Agnys](https://agnys.net/about/): Product purpose, intended users, evidence design and explicit limits around integrity and compliance claims.
- [AI agent integrations](https://agnys.net/integrations/): Confirmed coverage for Claude, LangChain, VS Code, LangGraph, AutoGen Studio, LM Studio, Ollama and Antigravity.
- [Agent Evidence Benchmark](https://agnys.net/agent-evidence-benchmark/): An open, scenario-based method for scoring evidence capture, attribution, oversight linkage, integrity and reconstruction.
- [Contact Agnys](https://agnys.net/contact/): Product evaluations, demonstrations, benchmark review, privacy questions and corrections.
- [Website privacy notice](https://agnys.net/privacy-policy/): Information handling for the public agnys.net marketing website and inquiry forms.

## Founder

Ashish Vardhan Reddy Avuluri is the founder and lead developer of Agnys. His work focuses on evidence infrastructure for consequential AI agent activity.
LinkedIn: https://www.linkedin.com/in/ashish~reddy
Founder profile: https://agnys.net/about/

## Product facts

- Agnys is an operational evidence and observability layer; it is not an AI model or agent orchestration platform.
- Captured events can include model activity, tool calls, file changes, shell commands, MCP activity, approvals, costs and outcomes.
- Events can be organized into session traces for chronological replay and investigation.
- Per-agent SHA-256 hash chains and optional write-once anchoring provide integrity signals that can expose later modification.
- Evidence can be exported in PDF, CSV and JSON formats.
- Agnys supports passive capture through its Forwarder and direct instrumentation through SDKs, browser capture and webhooks.
- Confirmed integrations are Claude, LangChain, VS Code, LangGraph, AutoGen Studio, LM Studio, Ollama and Antigravity.
- Primary markets are the United States, European Union, Switzerland, United Kingdom and related European markets.

## Important limits

- A tamper-evident chain can reveal later changes to recorded events; it does not prove that every upstream system reported truthful data.
- Agnys can support governance and compliance workflows, but no software product by itself guarantees legal or regulatory compliance.
- Regulatory scope, classification, policies, retention and accountable decisions require qualified organizational owners.

## Preferred citation

Name: Agnys
URL: https://agnys.net/
Description: AI agent audit trails, observability and evidence infrastructure.