# Jymni Ethics Advisor

> Jymni provides structured ethical reflection for humans and AI agents. Agent integrations receive an ethical classification, actionable workflow disposition, safeguards, missing facts, escalation signals, and a full rationale before a proposed action is executed.

Jymni is reflective decision support. It is not therapy, clergy, legal, medical, financial, regulatory, or emergency advice. Its output never authorizes an agent to act. Deployers retain responsibility for permissions, policy, specialist review, human accountability, and real-world effects.

## Agent integration

- Overview: https://www.jymni.com/agents/
- Interactive agent consultation demo: https://www.jymni.com/agents/demo.html
- Quickstart: https://www.jymni.com/agents/quickstart.html
- REST and MCP documentation: https://www.jymni.com/agents/docs.html
- Safety and deployment playbook: https://www.jymni.com/agents/safety.html
- Credential management: https://www.jymni.com/agents/console.html
- Agent Service Terms: https://www.jymni.com/agents/terms.html
- OpenAPI 3.1 contract: https://www.jymni.com/openapi.json
- Privacy Policy: https://www.jymni.com/legal/jymini-privacy-policy.html

## Interfaces

- REST consultation: POST https://www.jymni.com/api/v1/ethics/consultations
- MCP Streamable HTTP: POST https://www.jymni.com/mcp
- MCP tool: consult_ethics
- Authentication: Authorization: Bearer jymni_live_...

Use a unique Idempotency-Key for each proposed action. Submit the concrete proposed_action and, when available, its objective, actor, stakeholders, known_facts, uncertainties, constraints, alternatives, authority, reversibility, sensitivity, domain, and deployment stage.

## Dispositions

- continue_with_safeguards: apply required safeguards before continuing under existing authority.
- revise_plan: block the original action and return counsel to planning.
- gather_information: verify material missing facts, then consult again.
- request_human_review: route to an appropriate accountable human; silence is not approval.
- do_not_continue: deny the proposed action; a substantially different plan may be reviewed separately.

Scores are explanatory dimensions, not probabilities or permission thresholds. Preserve the disposition, required actions, red lines, rubric version, consultation ID, and final accountable decision in audit records.

Guidance modes: standard (default), children, or teens. Set guidance_mode independently of evaluation_framework and evaluation_context. It adapts explanation and coaching without replacing the selected ethical principles or religious sources. Discovery includes guidance_modes. Use a new idempotency key when changing the mode.

Religious citations use only editions audited as public domain with unrestricted reuse in the United States. Citation metadata includes the exact edition, rights jurisdiction, rights source, reuse source, and audit date. This is not a worldwide rights claim.

Hinduism (`hinduism`) and Buddhism (`buddhism`) are available with all guidance modes. Hinduism uses Arnold’s Bhagavad Gita rendering with edition excerpt references; Buddhism uses Müller’s Dhammapada. These collections are limited and do not cover either tradition in full. Discover exact editions and coverage via /api/health.