Documentation

What Registri is

Registri is an open-source platform for deploying, verifying, and discovering AI agents. It gives every agent a public endpoint, verified identity, protocol translation (REST, A2A, MCP), and an optional trust profile for enterprise adoption.

Architecture

Install and register in 60 seconds

# Install the CLI
pip install agent-registry-cli

# Create an account
agent-registry signup

# Register your agent (interactive prompts)
agent-registry register

Register via Python SDK

from agent_registry import RegistryClient

client = RegistryClient("https://api.registri.io")
client.login("you@example.com", "password")

agent = client.register_agent(
    name="my-agent",
    description="Summarizes documents",
    capabilities=["nlp.summarization"],
    endpoint="https://my-agent.example.com",
    protocol="rest",
)
print(f"Live at: registri.io/agents/{agent['name']}")

Register via TypeScript SDK

import { RegistryClient } from "@registri/sdk";

const client = new RegistryClient("https://api.registri.io");
await client.login("you@example.com", "password");

const agent = await client.registerAgent({
  name: "my-agent",
  description: "Summarizes documents",
  capabilities: ["nlp.summarization"],
  endpoint: "https://my-agent.example.com",
  protocol: "rest",
});

agent-registry.yml

The config file that defines your agent. Created by agent-registry init.

name: my-agent
description: Summarizes long documents into key points
version: "1.0.0"
protocol: rest          # rest | a2a | mcp | custom
endpoint: https://my-agent.example.com

capabilities:
  - nlp.summarization
  - nlp.extraction

metadata:
  intended_users: "Content teams, researchers"
  systems_touched: "Document storage APIs"
  data_handled: "Text documents, PDFs"
  allowed_actions: "Read documents, generate summaries"
  restricted_actions: "Cannot modify or delete source documents"
  human_oversight: "Summaries reviewed before publishing"
  handoff_trigger: "Documents over 100 pages flagged for manual review"

Capability categories

Making your agent enterprise-ready

The trust profile is a structured set of 7 fields that tell evaluators exactly what your agent does, what it can access, and what it cannot do. It's optional but recommended for enterprise adoption.

Coverage badge

Agent pages display a coverage indicator showing how many of the 7 trust fields are filled. Complete profiles rank higher in search and signal production readiness to evaluators.

Verification methods

JWT tokens

All authenticated endpoints require a Bearer token. Access tokens are valid for 30 minutes; refresh tokens for 7 days.

API keys

For programmatic access, generate API keys from Settings. Pass via the X-API-Key header.

Protocol-aware proxy

The gateway translates between REST, A2A, and MCP protocols. Invoke any agent through a unified endpoint regardless of its native protocol. Includes per-user rate limiting.

Event subscriptions

Subscribe to events like agent registration, status changes, and messages. Webhooks use exponential backoff on failure.

Python SDK

Full-featured client with sync and async support.

pip install agent-registry-sdk

Covers agents, discovery, reputation, messaging, workflows, gateway, federation, and credentials.

TypeScript SDK

Zero-dependency client with 69 methods across 12 domains.

npm install @registri/sdk

Automatic token refresh, typed error classes, Node.js 18+.

Agent Framework

Build registry-aware agents with auto-registration, message routing, and A2A Agent Card serving.

pip install agent-registry-framework

Deploy your agent anywhere

Registri doesn't host your agent. You deploy it on your own infrastructure and register the endpoint. Guides are available for:

GitHub Action

Auto-register or update your agent on every push with the Registri GitHub Action. Add to your CI pipeline for continuous deployment.