agno-integrations

name: agno-integrations description: | Integrate Agno agents with observability platforms, A2A protocol, and external services. Covers OpenTelemetry, Langfuse, Arize Phoenix, Agent-to-Agent protocol, and Discord bots. Trigger this skill when: adding observability to agents, setting up tracing, integrating with Langfuse or similar platforms, or asking "how do I monitor my agents?" license: Apache-2.0 metadata: version: "1.0.0" author: agno-team tags: ["observability", "tracing", "integrations", "langfuse", "a2a", "agno"]

Integrate Agno Agents

Connect agents to observability platforms, external services, and other agent systems. Install with pip install agno.

Observability with OpenTelemetry

Langfuse via OpenInference

import os
from opentelemetry import trace
from opentelemetry.sdk.trace import TracerProvider
from opentelemetry.exporter.otlp.proto.http.trace_exporter import OTLPSpanExporter
from opentelemetry.sdk.trace.export import SimpleSpanProcessor
from openinference.instrumentation.agno import AgnoInstrumentor

# Configure Langfuse OTLP endpoint
os.environ["OTEL_EXPORTER_OTLP_ENDPOINT"] = "https://cloud.langfuse.com/api/public/otel"
os.environ["OTEL_EXPORTER_OTLP_HEADERS"] = (
    f"Authorization=Basic {os.environ['LANGFUSE_AUTH']}"
)

# Set up tracing
provider = TracerProvider()
provider.add_span_processor(
    SimpleSpanProcessor(OTLPSpanExporter())
)
trace.set_tracer_provider(provider)

# Instrument Agno
AgnoInstrumentor().instrument()

# Now all agent runs are automatically traced
from agno.agent import Agent

agent = Agent(
    model="openai:gpt-4o",
    markdown=True,
)

agent.print_response("Hello!", stream=True)

Arize Phoenix (Local)

from opentelemetry import trace
from opentelemetry.sdk.trace import TracerProvider
from opentelemetry.exporter.otlp.proto.grpc.trace_exporter import OTLPSpanExporter
from opentelemetry.sdk.trace.export import SimpleSpanProcessor
from openinference.instrumentation.agno import AgnoInstrumentor

provider = TracerProvider()
provider.add_span_processor(
    SimpleSpanProcessor(
        OTLPSpanExporter(endpoint="http://localhost:6006/v1/traces")
    )
)
trace.set_tracer_provider(provider)

AgnoInstrumentor().instrument()

Required Dependencies

pip install "agno[opentelemetry]"
pip install openinference-instrumentation-agno
pip install opentelemetry-exporter-otlp

Agent-to-Agent (A2A) Protocol

Agno supports the A2A protocol for inter-agent communication across systems:

from agno.agent import Agent
from agno.tools.a2a import A2ATools

# Connect to a remote A2A-compatible agent
agent = Agent(
    model="openai:gpt-4o",
    tools=[A2ATools(url="http://remote-agent:8000/a2a")],
    instructions=["Use the remote agent for specialized tasks."],
)

agent.print_response("Ask the remote agent to analyze this data.", stream=True)

MCP (Model Context Protocol)

Connect to MCP servers for external tool access:

from agno.agent import Agent
from agno.tools.mcp import MCPTools

agent = Agent(
    model="openai:gpt-4o",
    tools=[MCPTools(url="http://localhost:3000/mcp")],
)

agent.print_response("Use the MCP tools to help me.", stream=True)

Slack Integration

from agno.agent import Agent
from agno.tools.slack import SlackTools

agent = Agent(
    model="openai:gpt-4o",
    tools=[SlackTools()],
    instructions=["Help manage Slack messages and channels."],
)

GitHub Integration

from agno.agent import Agent
from agno.tools.github import GithubTools

agent = Agent(
    model="openai:gpt-4o",
    tools=[GithubTools()],
    instructions=["Help manage GitHub issues and PRs."],
)

Email Integration

from agno.agent import Agent
from agno.tools.gmail import GmailTools

agent = Agent(
    model="openai:gpt-4o",
    tools=[GmailTools()],
    instructions=["Help manage emails."],
)

Custom Observability Hook

Log agent events programmatically:

from agno.agent import Agent

def on_run_complete(run_response):
    print(f"Run {run_response.run_id} completed")
    if run_response.metrics:
        print(f"  Tokens: {run_response.metrics.total_tokens}")
        print(f"  Duration: {run_response.metrics.time_to_first_token}s")

agent = Agent(
    model="openai:gpt-4o",
    markdown=True,
)

response = agent.run("Hello!")
on_run_complete(response)

Anti-Patterns

Don't instrument in production without sampling — tracing every request adds overhead
Don't hardcode auth tokens — use environment variables for Langfuse/Arize keys
Don't skip AgnoInstrumentor().instrument() — it must be called before agent runs
Don't mix tracing backends — pick one observability platform per deployment
Don't forget to install extras — agno[opentelemetry] is required for tracing