🌊 Stream-First Architecture

Watch agents
think in real-time

Stream-first architecture where agents yield execution steps for transparent reasoning. See every thought, decision, and action as it happens.

See Live Streaming ← Back to Home

Real-Time Agent Execution

Watch every step unfold as your agent works

streaming_agent.py 
 from cogency.agent import Agent from cogency.llm import GeminiLLM agent = Agent( name="StreamAgent", llm=GeminiLLM(api_key="your-key") ) # Stream the agent's execution in real-time async for chunk in agent.stream("What is 127 * 43?"): print(chunk)

Live Agent Output

🧠 PLAN: I need to calculate 127 * 43. I'll use the calculator tool.
💭 REASON: This is a simple multiplication that I can solve with tools.
⚡ ACT: Using calculator tool with 127 * 43...
🔄 REFLECT: Calculator returned 5461. This seems correct.
✨ RESPOND: The answer is 5461.

Rich Stream Types

Execution Streams

  • 🧠 Planning: Goal decomposition and step planning
  • 💭 Reasoning: Analysis and decision-making process
  • ⚡ Actions: Tool calls and external interactions
  • 🔄 Reflection: Self-evaluation and error correction

Monitoring Streams

  • 📊 Metrics: Performance and usage statistics
  • 🔍 Traces: Full execution traces with timing
  • ⚠️ Errors: Exception details and stack traces
  • ✅ Events: Lifecycle and state change events

Why Streaming Matters

Full Transparency

See exactly how your agent reasons through problems, making debugging and optimization straightforward.

Instant Feedback

No waiting for final results—see progress immediately and intervene when needed for better user experience.

Production Monitoring

Built-in observability for production deployments with real-time performance metrics and alerting.

Streaming Excellence

Multi-step Reasoning

Stream each step of the reasoning process live.

Production Ready

Real-time monitoring and performance metrics.

Zero Configuration

Streaming works out of the box, no setup required.

Ready for Real-Time Agents?

Experience the transparency and control of streaming AI execution

Start Streaming View on GitHub