Documentation

What is AgentMD?

New to agentic AI? What is Agentic AI? — A plain-language primer on autonomous agents and why governance matters.

Why execute AGENTS.md? — Most teams treat it as read-only. AgentMD executes it.

agents.md defines the AGENTS.md format — a README for agents. AgentMD goes further: we parse it, validate it, and execute the commands it describes.

• Parse — Sections, commands, YAML frontmatter, directives
• Validate — Format, safety, recommended sections
• Execute — Build, test, lint with permission boundaries
• Score — 0–100 agent-readiness score

Guides

GitHub Quick Start

Connect your repo in 5 minutes: sign in with GitHub, add a repository, ensure AGENTS.md exists, run your first execution. Install the GitHub App for repo discovery and webhooks.

Template Gallery

AGENTS.md templates for React, Next.js, Python, Rust, Go, and Java.

AI Governance & AgentOps

AgentMD supports agentic AI governance through guardrails, permissions, and policies. As regulations like the EU AI Act evolve, AgentMD helps you run agents safely and accountably. See Agentic AI Best Practices for IBM 2026 guidance (observable, adaptive, accountable).

• Guardrails — Declare in YAML frontmatter: "Never modify production", "Never merge", etc.
• Permissions — Explicit allowlists for shell commands, pull requests, and other resources.
• Policies — Block, warn, or require approval for agent actions (Ops dashboard).
• Audit — Execution history and audit logs for traceability.

AgentOps (IBM) — Lifecycle management for AI agents (development, testing, monitoring, feedback, governance). AgentMD complements observability tools by focusing on execution and governance. The page also links to the IDC whitepaper on evolving regulations and agentic AI.

• AI agent governance — Autonomy, transparency, compliance, and security risks
• AI agent ethics — Ethics risks, alignment, and function-calling hallucination
• AI agent evaluation — Metrics, benchmarks, and evaluation process (tutorial)
• AI agent security — Threat landscape, prompt injection, and best practices
• Human-in-the-loop tutorial — LangGraph HITL with static and dynamic interrupts
• AI agent observability — Langfuse with watsonx Orchestrate for traces and metrics

Recommended Agent Skills

Agent skills extend AI coding assistants with specialized knowledge. Install these to improve AGENTS.md workflows:

• getsentry/agents-md — Generate and manage AGENTS.md files. Use when creating or editing AGENTS.md.
  git clone --depth 1 --filter=blob:none --sparse https://github.com/getsentry/skills.git && cd skills && git sparse-checkout set plugins/sentry-skills/skills/agents-md && cp -r plugins/sentry-skills/skills/agents-md ~/.cursor/skills/
• agentmd/agentmd — AgentMD CLI execution, validation, and scoring. Use when running agentmd commands, validating AGENTS.md, or computing scores.
  cp -r skills/agentmd ~/.cursor/skills/

  From repo root. Or clone from GitHub and copy the skills/agentmd folder.

awesome-agent-skills — Browse 380+ skills from Vercel, Stripe, Sentry, etc. Compatible with Cursor, Codex, Claude Code, and Gemini CLI.

Skills are curated, not audited. Review sources before use.

Integrations

AgentMD integrates with Slack, Jira, and GitHub:

• Slack — Human-in-the-loop approvals. Configure SLACK_SIGNING_SECRET and the Interactivity URL.
• Jira — Post execution status on completion. Set JIRA_WEBHOOK_URL to receive success/failure payloads.
• GitHub — Repository discovery, PR checks, webhook-triggered runs.

See docs/INTEGRATIONS.md for setup details.

Governance Roadmap

Planned features to support evolving AI regulations and risk management:

• EU AI Act risk classification and compliance workflows
• Enhanced traceability and decision audit trails
• Automated risk assessment for agent configurations
• OpenTelemetry export for Langfuse, Datadog, and observability platforms
• Emergency pause / kill switch for running executions
• Integration with external governance platforms