> For the complete documentation index, see [llms.txt](https://docs.metamynd.ai/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.metamynd.ai/introduction/welcome.md).
# Welcome
Docs for MetaMynd: give AI agents a verifiable identity, a mandate with limits, and a gate that checks every action and records signed evidence.
Metamynd is the trust layer for autonomous AI agents. This documentation covers **AgentSafe**, our live governance product, and the two systems underneath it: **MAGP** (the Metamynd Agentic Governance Protocol) and **VeriFAI** (identity, ownership, and credentialing).
If you're new here, the fastest way in is:
1. Read [The Metamynd Thesis](https://docs.metamynd.ai/introduction/the-metamynd-thesis) and [Mission](https://docs.metamynd.ai/introduction/mission). Two pages, five minutes, and you'll understand *why* this exists, which makes every reference page below make more sense.
2. Run the [Quickstart](https://docs.metamynd.ai/introduction/quickstart). One command scaffolds a governed agent and a separate gateway process on the free testnet tier, with no verification needed to start. There is a no-account option too, and a Python path.
3. From there, branch into [Core Concepts](https://docs.metamynd.ai/core-concepts/the-agentic-economy-and-the-trust-gap) if you want the mental model first, or straight into the [Agent Protocol](https://docs.metamynd.ai/agent-protocol/protocol-overview) reference if you're integrating today.
#### **What Metamynd actually is**
One sentence: **the agent can ask, only the gateway can act.**
An AI agent signs a request describing what it wants to do. It never holds the real credential (the payment method, the upstream API key, the database write permission) that would let it act directly. The MAGP gate decides whether the request is within the agent's policy. The gateway that holds the real credential then re-verifies the signed request for itself and **claims** the authorization before it acts.
That claim is single-use: of any number of executors, exactly one can claim a given authorization, so a permitted action runs once and a replay goes nowhere. The claim also checks that what the gateway is about to execute is exactly the payload the agent signed, and on mainnet only a gateway the owner has registered as a trusted counterparty can make it.
That's the entire architecture. Everything else in these docs is how it is enforced, made trustless, and made auditable.
#### **What's live vs. what's roadmap**
We mark every capability in this documentation with one of three states, because a governance product that oversells its own maturity is not one you should trust:
| | |
| ------------------ | -------------------------------------------------------------------------------------------------- |
| 🟢 **Live** | Running in the reference deployment at [metamynd.ai](https://metamynd.ai) today. |
| 🟡 **In progress** | Implemented and demonstrated end-to-end (usually in the reference demo), rolling out more broadly. |
| ⚪ **Roadmap** | Designed, not yet built. |
#### **Where things live**
| Product | What it does |
| ------------- | --------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- |
| **AgentSafe** | The governance plane: no-code policy engine, the MAGP authorize gate, evidence ledger, escalation queue, and settlement of every value-bearing action. |
| **VeriFAI** | The identity plane: `did:hedera` and `did:key` identities on testnet and mainnet (mainnet requires a KYC/KYB-verified principal), delegated authority, and credential custody through the Credential Vault and the local signer daemon. |
| **MAGP** | The open wire protocol both products speak: signed requests, deterministic policy evaluation, cross-party re-verification, and exactly-once execution through single-use claims, payload binding and trusted counterparties. |
#### **Developer surfaces**
| Surface | What it's for |
| ---------------------------------------------------------------------------------------------------- | -------------------------------------------------------------------------------------------------------------- |
| [`create-metamynd-agent`](https://www.npmjs.com/package/create-metamynd-agent) | Scaffold a governed agent plus its gateway in one command (testnet). |
| [`@metamynd/agentsafe-guard`](https://www.npmjs.com/package/@metamynd/agentsafe-guard) | Agent-side guard: sign, authorize, and gate tools. |
| [`@metamynd/agentsafe-mcp-guard`](https://www.npmjs.com/package/@metamynd/agentsafe-mcp-guard) | Service-side guard for MCP servers: re-verify, claim, settle. |
| [`@metamynd/agentsafe-http-gateway`](https://www.npmjs.com/package/@metamynd/agentsafe-http-gateway) | A governed reverse proxy for any HTTP route, with payload binding and Credential Vault support. |
| [`@metamynd/agentsafe-a2a-guard`](https://www.npmjs.com/package/@metamynd/agentsafe-a2a-guard) | Governance for the Agent2Agent (A2A) protocol. |
| [`@metamynd/agentsafe-signer`](https://www.npmjs.com/package/@metamynd/agentsafe-signer) | A local signer daemon, so agent and service keys never enter the agent's process. |
| [`metamynd-client`](https://pypi.org/project/metamynd-client/) (PyPI) | The Python client, with runnable examples for LangGraph, CrewAI, OpenAI Agents SDK, Pydantic AI and LangChain. |
| [Launchpad](https://metamynd.ai/dashboard/launchpad) (dashboard) | Launch mainnet agents once your owner verification is complete. |
| [MAGP v1.0 specification](https://metamynd.ai/developers/spec) | The normative protocol. |
---
# Agent Instructions
This documentation is published with GitBook. GitBook is the documentation platform designed so that both humans and AI agents can read, navigate, and reason over technical content effectively. Learn more at gitbook.com.
## Querying This Documentation
If you need additional information that is not directly available in this page, you can query the documentation by asking a question.
Perform an HTTP GET request on the following URL with the `ask` and `goal` query parameters:
```
GET https://docs.metamynd.ai/introduction/welcome.md?ask=&goal=
```
`ask` is the immediate question: it should be specific, self-contained, and written in natural language.
`goal` is what the user is ultimately trying to achieve, the reason they need the answer. Sharing it helps GitBook give you a better, more relevant answer. A goal is most helpful when it describes the outcome the user wants rather than restating the question. For example, with `ask=how do I create an API token`, a goal like `automate deployments from our CI pipeline` lets GitBook tailor the answer to that use case.
The response will contain a direct answer to the question and relevant excerpts and sources from the documentation.
Use this mechanism when the answer is not explicitly present in the current page, you need clarification or additional context, or you want to retrieve related documentation sections.
