agentactionprotocol

agentactionprotocol / aap

Public

AAP: Agent Action Protocol — a portable, composable spec for AI agents taking action in the world

11
0
69% credibility
Found Mar 07, 2026 at 11 stars -- GitGems finds repos before they trend. Get early access to the next one.
Sign Up Free
AI Analysis
TypeScript
AI Summary

AAP is a draft open-source specification for defining, securing, and observing actions performed by AI agents in a portable and composable way, including an adapter example for a specific runtime.

How It Works

1
📖 Discover AAP

You hear about AAP, a helpful guide that makes AI helpers safer when they do real things like send messages or check files.

2
💡 Understand the idea

You read the simple explanation of how to describe what your AI can do, with built-in safety checks and tracking.

3
✨ Describe your AI action

You create a short list in plain words of what your AI action needs, produces, and how much it can spend.

4
🛡️ Add safety rules

You set gentle limits like who can approve it, cost caps, and ways to watch what happens.

5
🔗 Connect to your AI setup

You easily link these descriptions to your AI agent's world so it follows the rules every time.

6
👀 Watch it work safely

You see your AI check everything before acting, log what happens, and stay within safe bounds.

🎉 Reliable AI actions

Your AI helper now takes actions confidently, with full control, recovery if needed, and peace of mind.

Sign up to see the full architecture

5 more

Sign Up Free

Star Growth

See how this repo grew from 11 to 11 stars Sign Up Free
Repurpose This Repo

Repurpose is a Pro feature

Generate ready-to-use prompts for X threads, LinkedIn posts, blog posts, YouTube scripts, and more -- with full repo context baked in.

Unlock Repurpose
AI-Generated Review

What is aap?

AAP is a TypeScript spec for defining AI agent actions via portable YAML manifests, handling everything from inputs and permissions to checkpointing and cost limits. It solves the chaos of agents executing real-world tasks like emailing or deploying code, where protocols like MCP only cover transport, leaving gaps in contracts, composability, and observability. Developers get a "package.json for actions" that works across runtimes via adapters, like the OpenClaw one for hooking in OpenTelemetry traces.

Why is it gaining traction?

Unlike MCP's transport focus, AAP layers on action identity, lifecycle hooks for auth and guards, and DAG compositions with budget propagation—making agents safer and more portable without runtime lock-in. The OpenClaw adapter adds manifest validation and context injection with zero breaking changes, letting you wrap handlers for instant contract enforcement and tracing. Early adopters dig the declarative approach for production agents, dodging hardcoded auth and unchecked costs.

Who should use this?

Agent builders on OpenClaw or MCP stacks integrating tools like Anthropic or LangChain, especially ops teams enforcing budgets on actions like ticket creation or deployments. Suited for devs standardizing agent pipelines across CLIs, REST, or AI models, avoiding per-runtime rewrites.

Verdict

Draft v0.1 with 11 stars and 0.7% credibility score signals raw potential but early maturity—solid docs and examples, though needs runtime buy-in for traction. Try the OpenClaw adapter if you're prototyping observable agents; contribute to shape the spec.

(178 words)

Sign up to read the full AI review Sign Up Free

Similar repos coming soon.