We can't find the internet
Attempting to reconnect
Something went wrong!
Hang in there while we get back on track
Connect agents from LangChain, AutoGen, CrewAI, and any framework. Task queues, team coordination, semantic matching, and enterprise-ready monitoring—all in one platform.
Enterprise-grade capabilities that transform how you build and deploy multi-agent systems
Queue and manage long-running tasks with automatic retries, priority handling, and real-time status tracking. Perfect for async workflows.
NEWOrganize agents into teams with roles. Broadcast messages to entire groups and coordinate complex multi-agent workflows.
NEWAgents never miss a message. Automatic queuing for offline agents with instant delivery when they reconnect.
NEWFind agents using natural language queries. Powered by AI embeddings—describe what you need, we find the right agent.
AIReal-time insights into message volume, agent activity, costs, and performance. Visual dashboards with customizable date ranges.
NEWConfigure custom alerts for usage thresholds, agent downtime, error rates, and latency. Notifications via email, webhook, or Slack.
NEWSecurity-compliant logging of all agent interactions, messages, and system events. Export for compliance and debugging.
NEWMulti-user organizations with role-based access control. Invite team members, manage API keys, and track per-user activity.
NEWSee how teams are using CACP to build sophisticated multi-agent systems
LangChain agents analyze code, AutoGen agents run tests, and a reporting agent compiles results—all coordinated via task queues.
Tier 1 agents triage, specialists solve complex issues, and semantic matching routes to the right expert automatically.
CrewAI research agents collaborate, sharing findings through group broadcasts. Offline agents queue results for later analysis.
Alerts trigger agent teams to investigate. Logs are audited, teams coordinate via groups, and status updates flow in real-time.
Data ingestion agents queue tasks, analysis agents process in parallel, and analytics dashboards track costs and throughput.
Agents discover each other via semantic matching for language-specific tasks. Group broadcasts coordinate project-wide updates.
Why developers and organizations choose CACP for their multi-agent systems
Your agents work with any framework. Switch from LangChain to AutoGen without rewriting code. Your infrastructure, your choice.
Built on Erlang/OTP—the same tech powering 90% of global telecoms. 99.99% uptime guaranteed with automatic failover.
Rate limiting, usage quotas, real-time analytics, audit logs, and team management built-in. Monitor every message.
Connect your first agent in under 5 minutes. Clean APIs, clear docs, and SDKs for Python, JavaScript, and Go.
Real-time WebSocket for live interactions. Batch HTTP for bulk operations. One broker handles both seamlessly.
Transparent protocol specification. No hidden APIs. Join a community building the standard for agent communication.
Technical capabilities that enable seamless agent-to-agent communication
Agents advertise their skills and discover compatible peers automatically. No manual configuration needed.
Use natural language to find the right agent. AI-powered semantic matching understands context and intent.
Group broadcasts and team-based routing. Organize agents into squads for complex, coordinated workflows.
Offline agents never miss a message. Intelligent queuing with instant delivery on reconnection.
Queue long-running tasks with priorities and automatic retries. Perfect for async, multi-step workflows.
Multiple frameworks, one protocol
LangChain
Research
CACP
Orchestrator
AutoGen
Analysis
Use the shipped HTTP and Phoenix Channel contracts with public status lookup
import asyncio
import json
import websockets
async def run():
token = "YOUR_AGENT_JWT"
uri = "ws://localhost:4001/ws/v1/websocket?token={token}&vsn=2.0.0"
async with websockets.connect(uri) as ws:
await ws.send(json.dumps({
"topic": "agent:my-agent-123",
"event": "phx_join",
"payload": {"token": token},
"ref": "1"
}))
await ws.send(json.dumps({
"topic": "agent:my-agent-123",
"event": "send",
"payload": {"message": {
"message_id": "550e8400-e29b-41d4-a716-446655440000",
"sender": {"agent_id": "my-agent-123"},
"recipient": {"agent_id": "other-agent", "discovery": "direct"},
"message_type": "task",
"payload": {"description": "Analyze this code for bugs"}
}},
"ref": "2"
}))
print(await ws.recv())
asyncio.run(run())
Support for all major AI agent frameworks with more on the way
LangChain
Python & JavaScript
AutoGen
Microsoft
CrewAI
Python
Vertex AI
Google Cloud
AgentOps
Observability
GPT Researcher
Research Agents
Custom Agents
Any Framework
More Coming
Stay Tuned
All features available to everyone. No credit card required. No limits on creativity.
🎉 CACP is now 100% free for everyone. All plans include full access to all features.
Everything you need to know about CACP
Everything you need to know about CACP and AI agent communication
CACP (Cross-Agent Communication Protocol) is a universal messaging and RPC layer that enables AI agents built on different frameworks to communicate. Think of it as TCP/IP for AI Agents - an open protocol specification with commercial hosted infrastructure.
CACP works with LangChain, AutoGen, CrewAI, Vertex AI, Claude Agents, and any custom agent implementation. You can connect agents from any framework using our HTTP REST API or WebSocket channels.
Yes! CACP is 100% free forever. No credit card required. All features are available to everyone including task queues, semantic discovery, analytics, alerts, and team management. We believe in open protocols for the AI community.
Connect your first agent in under 5 minutes. Use our Python, JavaScript, or Go SDKs. Simply register your agent, get a JWT token, and start sending messages via HTTP or WebSocket. Check our documentation for step-by-step guides.
Semantic discovery uses AI embeddings to find agents using natural language queries. Instead of exact capability matching, describe what you need (e.g., 'find an agent that can analyze PDFs') and CACP finds the right agent automatically.
Absolutely! CACP is enterprise-ready. Built on Erlang/OTP (the same tech powering 90% of global telecoms) with 99.99% uptime guarantee. Features include rate limiting, audit logs, data retention policies, team management with RBAC, and SOC 2 / GDPR compliance.
CACP provides 4 priority levels (critical, high, normal, low) for task queuing with automatic retries, status tracking, and async handling. Perfect for long-running workflows that need reliable delivery. Failed tasks are automatically retried with exponential backoff.
Offline agents never miss a message. CACP automatically queues messages and delivers them instantly when agents reconnect. Guaranteed delivery with intelligent persistence - your agents stay synchronized even through disconnections.
Still have questions?
Join thousands of teams building the future of AI agent communication. Task queues, team coordination, semantic matching—it's all here.
100% Free Forever • No credit card required • Enterprise-ready