NOW OPEN SOURCE — v0.1 LIVE

AIP

Artificial Intelligence Protocol

We built the communication standard for the AI agent internet. AIP is to AI agents what HTTP was to the web — a universal protocol so any agent can talk to any other agent, regardless of who built them.

View on GitHub → How It Works

⚡ Canner AI Solutions Releases AIP v0.1

Today we're open-sourcing AIP — a signed, versioned, capability-aware protocol for AI agents to discover each other, authenticate, delegate tasks, stream progress, and return results. Built from the ground up as infrastructure, not a product. The goal is a standard the entire AI ecosystem can build on.

// Architecture

The AI Infrastructure Stack

05 AI Operating Layer FUTURE

04 AI Memory Network FUTURE

03 AIP — Agent Protocol YOU ARE HERE

02 Internet / Cloud EXISTING

01 Hardware EXISTING

AIP is the communication layer. Memory and orchestration stack on top of it. We're building this from the ground up — the same way TCP/IP was built before the web existed.

// Protocol

How AIP Works

Identity

Every agent gets a cryptographic identity — a DID and Ed25519 keypair. Every message is signed. You always know exactly who you're talking to.

Handshake

Agents introduce themselves with HELLO, exchange capabilities and public keys, then complete a signed challenge-response auth before any work begins.

Task Delegation

One agent sends TASK_CREATE with typed inputs. The worker accepts, streams live status updates, and returns a structured result with a confidence score.

Full Auditability

Every message carries a trace ID and timestamp. Every signature is verifiable. Non-repudiation built in — enterprise-grade from day one.

aip_message.json

{
  "aip_version": "0.1",
  "type": "TASK_CREATE",
  "message_id": "3f2e1a9b-c4d5-...",
  "from": "did:aip:canner:intake-agent",
  "to": "did:aip:canner:strategy-agent",
  "timestamp": "2026-05-23T10:00:00Z",
  "sig": "TBimjRxx77qM4ySS1pccqU0zR7nz...",
  "payload": {
    "task_type": "consulting.automation_analysis",
    "title": "Analyze automation for Smile Dental Group",
    "inputs": {
      "tools": ["Dentrix", "Weave"],
      "pain_points": ["no-shows", "manual recall"]
    }
  }
}

// Specification

The 9 Message Types

HELLO

Open connection, advertise capabilities + public key

HELLO_ACK

Accept connection, return own capabilities

AUTH_CHALLENGE

Send a random nonce to sign

AUTH_RESPONSE

Return signed nonce — proves private key ownership

TASK_CREATE

Delegate a typed task with structured inputs

TASK_ACCEPT

Confirm receipt, estimate duration

TASK_STATUS

Stream live progress updates (1–N per task)

TASK_RESULT

Final structured output + confidence score

ERROR

Structured failure with error code + retry signal

// Roadmap

What We're Building

v0.1

Task Delegation Protocol LIVE

Ed25519 signed envelopes, HELLO/AUTH handshake, full task lifecycle, WebSocket streaming, JSON Schema validation.

v0.2

Capability Registry

Agent discovery by capability name. No hardcoded URLs. Federated registry architecture.

v0.3

AI Memory Layer

Agents write and semantically retrieve shared knowledge. Memory provenance + confidence scoring.

v1.0

Federated Agent Network

DHT-based peer discovery, no central server. Cross-namespace trust. Binary encoding for performance.

AIP