As AI agents get smarter and more autonomous, we’re trusting them to book travel, negotiate deals, and access sensitive data.
But here’s the problem: there’s still no secure way to prove who an agent is — or who authorized it.
Let’s fix that.
Meet MCP-I: Model Context Protocol – Identity
MCP-I is an open specification that brings identity and delegation to AI agents. It extends the Model Context Protocol by letting agents and humans cryptographically verify each other before interacting.
Think of it as an identity layer for the AI agent ecosystem — built with did‘s and Verifiable Credentials, but abstracted enough to be developer-friendly.
💡 Why This Matters
We learned the hard way with email: protocols were built without trust or authentication, and we’ve been backfilling security ever since..
We don’t want to make the same mistake with agents.
Today, any actor can impersonate an agent, forge delegation, or spoof access to APIs — and there’s no standard way to verify what’s real.
MCP-I aims to change that.
The Building Blocks
- Decentralized Identity for Agents
- Delegation credentials issued by real users
- Edge enforcement using lightweight proxies (e.g. Cloudflare Workers)
- Zero backend changes to existing APIs
Supporting Tools (Built to Spec)
We’re releasing a set of tools to help teams adopt MCP-Identity fast:
MCP-I Identity Server
A turnkey MCP server for generating DIDs for users/agents, issuing credentials, managing delegation lifecycles, and verifying them on the edge (or with MCP-I aware services).
KnowThat.ai – Agent Reputation Directory
A public lookup service for AI agents and their delegations. Like credit scoring — but for autonomous software. Verify, flag, and query any agent.
Learn More
📖 MCP-I Spec: https://modelcontextprotocol-identity.io
👨💻 Join Us
We’re looking for:
-
Engineers building with agent frameworks (LangChain, AutoGPT, etc.)
-
Contributors to open identity standards
-
Feedback from devs working on agent infrastructure
Let us know what you think — drop a comment, DM, or just follow along. Let’s build a future where agents are not just intelligent — but trustworthy.
Tags: #aiagents #identity #weboftrust #security #openstandards #mcp #vouched
