# Mohar

> Open-source trust layer for AI agents. Authority is earned per capability from verified outcomes and lost immediately on failure. Every decision produces a signed, hash-chained receipt. Early stage: the design is specified, implementation is in progress.

## Docs

- [Design](https://mohar-liard.vercel.app/docs/design.md): How Mohar decides what an agent may do, how trust is earned and lost, and what each receipt proves.
- [Roadmap](https://mohar-liard.vercel.app/docs/ROADMAP.md): What gets built, in what order.
- [Integrations](https://mohar-liard.vercel.app/docs/integrations.md): How Mohar plugs into Claude Code, Codex, CAO, ADK, MCP, A2A and agent governance platforms, and how strong each seam is.
- [Sustainability](https://mohar-liard.vercel.app/docs/business-model.md): What stays free, what could be paid for later, and what would trigger a company.
- [Engineering bar](https://mohar-liard.vercel.app/docs/engineering-bar.md): The correctness, reliability, performance, security and compatibility standard every part of Mohar is held to.
- [Compliance](https://mohar-liard.vercel.app/docs/compliance.md): How Mohar maps onto the EU AI Act, GDPR, the Cyber Resilience Act, US export rules and NIST, SOC 2 and ISO frameworks.

## Repository

- [Source and issues](https://github.com/harshitwandhare/mohar): start with AGENTS.md if you are a coding agent.