Backed byCombinator

Debug Agents

Branch, compare, and merge AI agent executions with time-travel debugging. The developer platform for the agentic future.

Simple Integration

Drop-in observability for LangChain

Add a single callback handler to your existing agent executor and get instant visibility into every step, thought, and action.

Zero Config

Auto-tracks tools, LLM calls, and chain of thought.

Branching

Create new test branches from any step in history.

langchain_gemini.py

from realm import RealmClient, RealmCallbackHandler
from langchain.agents import AgentExecutor

# Initialize Realm
realm = RealmClient(project_id = "...")
session_id = realm.create_session(name = "Support Agent")

# Add tracking to your agent
agent_executor = AgentExecutor(
  agent = agent,
  tools = tools,
  callbacks = [RealmCallbackHandler(  # 👈 Just add this!
    client = realm,
    session_id = session_id
  )]
)

# Run as normal
agent_executor.invoke({ "input": "Help user123" })

Everything you need to build reliable agents

Move beyond console logs and trace viewers. Realm gives you full control over your agent's execution history.

Time-Travel Debugging

Replay any agent state at any point in time. Step forward and backward through execution history.

Step 12/45

Branching

Test alternative prompts or tools side-by-side. Create new test branches from any step.

Safe Merges

Automated conflict resolution ensures your improvements don't break functionality.

Zero Config

Just add the callback handler. We auto-track tools, LLM calls, and state.

Ready to ship agents with confidence?

Join the thousands of developers using Realm to debug, evaluate, and improve their AI agents.